A display module optical performance automatic testing method

Abstract

The application relates to the technical field of electric digital data processing, and particularly discloses a display module optical performance automatic testing method, which comprises the following steps: determining a feature vector set from an acquired image; calculating feature amplitudes in the feature vector set, and counting the number of feature amplitudes exceeding a preset deviation value; if the number exceeds a preset threshold value, determining a uniformity evaluation score; according to the uniformity evaluation score and the feature vector set, performing classification processing on a test picture to obtain an abnormal category; according to the abnormal category, a luminance gradient feature and a color deviation feature, adjusting the position parameters of the display module, and re-acquiring the test picture through an acquisition device; acquiring luminance uniformity values and contrast uniformity values from the test picture, and jointly calculating the comprehensive performance values; and generating a final test report. The display module optical performance automatic testing method solves the problems of low accuracy and reliability of the current testing method.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, and more specifically, to an automated testing method for the optical performance of display modules. Background Technology

[0002] In display modules, whether for smartphones, televisions, or automotive screens, display quality directly impacts user experience and product competitiveness. However, with the continuous expansion of display device types and application scenarios, the requirements for their quality testing are also increasing, and traditional testing methods are no longer sufficient to meet the needs of industry development. Against this backdrop, the technical challenges in the field of display quality testing are gradually becoming apparent.

[0003] The core issue is the precise control of the testing process. Display modules require specific visual displays to assess their performance during testing. If the device's position is not precisely adjusted, the collected information will be distorted, affecting the final judgment. For example, on a production line, a screen might display uneven colors in the test results due to a deviation in the viewing angle, even though the device itself is not faulty. This reduces the accuracy of the judgment, and such misjudgments directly impact production efficiency and product reliability. Summary of the Invention

[0004] In order to overcome the defects of the existing technology, the present invention provides an automated testing method for the optical performance of display modules, which aims to solve the problems in the existing technology.

[0005] The technical solution adopted by this invention to solve its technical problem is: an automated testing method for the optical performance of a display module, comprising the following steps: S1: Acquire the test screen of the display module to be tested as the acquired image; S2: Extract brightness gradient features and color deviation features from the acquired images to determine the feature vector set; S3: Calculate the feature amplitude of each two-dimensional feature in the feature vector set, and count the number of feature amplitudes that exceed the preset deviation value; if the number exceeds the preset threshold, perform region segmentation processing on the acquired image, calculate the brightness uniformity index of the segmented region, determine the uniformity evaluation score, and then execute step S4; otherwise, execute step S6. S4: Based on the uniformity evaluation score and feature vector set, a classification algorithm is used to classify the test screen of the display module, and the corresponding abnormal category is obtained from the classification result; S5: Based on the anomaly category, brightness gradient characteristics, and color deviation characteristics, adjust the position parameters of the display module, and re-acquire the test image of the display module to be tested through the acquisition device; S6: Obtain brightness and contrast characteristics from the test screen of the display module to be tested, calculate the corresponding brightness uniformity value and contrast uniformity value, and use a multi-parameter fusion processing method to jointly calculate the brightness uniformity value and contrast uniformity value to obtain the comprehensive performance value. S7: Determine whether the comprehensive performance value meets the preset standard, output the judgment result, and generate the final test report by combining the comprehensive performance value and the judgment result.

[0006] Preferably, in step S2, the gradient Gx in the horizontal direction and the gradient Gy in the vertical direction of each pixel are calculated on the acquired image based on the Sobel operator, where Gx represents the intensity of the horizontal brightness change and Gy represents the intensity of the vertical brightness change; the gradient magnitude M is obtained from Gx and Gy. ; By statistically analyzing the gradient magnitude M of each pixel in the entire acquired image, a set of brightness gradient features is obtained. The brightness gradient feature set is subjected to min-max normalization, and all gradient magnitudes M are linearly scaled to between 0 and 1 to obtain the brightness gradient features.

[0007] Optionally, in step S2, the acquired image is converted from the RGB color space to the Lab color space to obtain a Lab image; In the Lab image, standard whiteboard data is used as the standard color reference, and the Euclidean distance d between the Lab value of each pixel and the standard color reference is calculated as the color difference intensity. The pixel color difference calculation results are subjected to maximum and minimum normalization processing, and all color difference intensities d are linearly scaled to between 0 and 1 to obtain color deviation characteristics; The brightness gradient feature and color deviation feature of each pixel are sequentially concatenated into a two-dimensional feature. Then, all two-dimensional features are arranged according to the spatial arrangement of the pixels to encapsulate a set of feature vectors.

[0008] Specifically, in step S3, the feature magnitude z of each two-dimensional feature is calculated using the modulus formula, where the modulus formula is: , where x represents the brightness gradient feature and y represents the color deviation feature, thus obtaining the amplitude set; The number of two-dimensional features whose feature amplitude z exceeds a preset deviation value is counted from the amplitude set. If the number exceeds a preset threshold, the U-Net model is applied to the acquired image for region segmentation. The U-Net model takes the acquired image as input and outputs a segmentation mask. The list of boundary coordinates is extracted from the segmentation mask using the findContours function, resulting in a set of regions; For each region in the region set, the brightness gradient features of all pixels in the region are extracted, summed, and then divided by the number of pixels to obtain the average brightness of the region. Then, the brightness variance value of each region is calculated. The uniformity assessment score is obtained by summing all the brightness variance values ​​and dividing by the number of regions.

[0009] It is worth noting that in step S4, multiple sets of display module test data are obtained from the historical database. Each set of data includes labels for the abnormal category, uniformity evaluation scores, brightness gradient features of each pixel, and color deviation features of each pixel. Based on the test data, the uniformity evaluation score is used as the first feature; The average value of the brightness gradient features of all pixels is calculated to obtain the second feature; The average value of the color deviation features of all pixels is calculated to obtain the third feature; The first feature, the second feature, and the third feature are combined into a joint input vector; The label of the anomaly category is used as the target output, which together with the joint input vector constitutes a classification training sample set; Using the aforementioned classification training sample set, a multi-class support vector machine model is trained with a one-to-one residual strategy. The model outputs the confidence score of the joint input feature vector belonging to each preset anomaly category.

[0010] Preferably, in step S4, a uniformity evaluation score for the acquired image is obtained; Calculate the average value of the brightness gradient features of all pixels in the feature vector set of the acquired image; Calculate the average value of the color deviation features of all pixels in the feature vector set of the acquired image; The average value of the uniformity evaluation score, the average value of the brightness gradient feature, and the average value of the color deviation feature of the acquired image are used to construct the current input vector; Input the current input vector into the trained multi-class support vector machine model; The multi-class support vector machine model uses a decision function method to calculate the confidence score of the current input vector for each anomaly category; The anomaly category corresponding to the highest confidence score is obtained as the output category of the acquired image.

[0011] Optionally, in step S5, the average value of the brightness gradient feature and the average value of the color deviation feature are obtained and combined to form a normalized feature vector. For the labels of the aforementioned anomaly categories, the target vector of the standard module is determined by a lookup table. The target vector includes the corresponding standard values ​​of brightness gradient features and color deviation features. Based on the normalized feature vector and the target vector, the Euclidean distance between them is calculated as the fitness function; The position of the population is initialized using a differential evolution algorithm, where the position represents the orientation of the display module; Each position is evaluated according to the fitness function, and the optimal position is iteratively updated to minimize the fitness function, thus obtaining the final orientation of the display module. Based on the final orientation and the current orientation of the display module, optimized test configuration data is generated; The test configuration data is written into the driver chip of the display module, and the angle of the display module is adjusted to complete the orientation alignment correction. The test image displayed by the display module is obtained by acquiring the test screen through the acquisition device, and a resampled image is obtained.

[0012] Specifically, in step S6, the test screen of the display module to be tested is converted into a grayscale image, the grayscale value of each pixel is calculated for the grayscale image to obtain a brightness matrix, the standard deviation and the average value of the brightness matrix are calculated, and the standard deviation is divided by the average value to obtain the brightness uniformity value. Based on the brightness matrix, calculate the average difference between each pixel and all its neighboring pixels to obtain the contrast matrix; calculate the standard deviation and the average of the contrast matrix, and divide the standard deviation by the average to obtain the contrast uniformity value. Using preset weighting coefficients, the brightness uniformity value and the contrast uniformity value are weighted and summed to obtain the overall performance value.

[0013] The beneficial effects of this invention are as follows: In the automated testing method for the optical performance of display modules, a feature vector set is constructed by extracting brightness gradient and color deviation features from the acquired image, and the presence of acquisition angle deviation is determined by combining feature amplitude statistics. If a deviation is found, region segmentation is performed to accurately assess the uniformity of the image. A classification algorithm is then used to identify the anomaly category. Based on this, the display module position parameters are adjusted, the test image is re-acquired, and the brightness and contrast uniformity values ​​are fused to calculate the comprehensive performance value. Otherwise, the brightness and contrast uniformity values ​​of the currently acquired image are directly fused to calculate the comprehensive performance value, ultimately generating a test report. This solution effectively identifies the presence of acquisition angle deviation anomalies through multi-parameter fusion and dynamic adjustment strategies, and adjusts the position of the display module according to the anomaly type before re-acquiring, thereby improving the accuracy and reliability of display module testing. Attached Figure Description

[0014] Figure 1A flowchart for an automated testing method for the optical performance of display modules.

[0015] Figure 2 This is a flowchart of the steps in step S2.

[0016] Figure 3 This is a flowchart of the steps in step S4. Detailed Implementation

[0017] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0018] Combination Figures 1 to 3 An automated testing method for the optical performance of a display module, as shown, includes the following steps: S1: The test screen of the display module under test is captured as an image by a data acquisition device such as a camera; S2: Extract brightness gradient features and color deviation features from the acquired images to determine the feature vector set; S3: Calculate the feature amplitude of each two-dimensional feature in the feature vector set, and count the number of feature amplitudes that exceed the preset deviation value; if the number exceeds the preset threshold, perform region segmentation processing on the acquired image, calculate the brightness uniformity index of the segmented region, determine the uniformity evaluation score, and then execute step S4; otherwise, execute step S6. S4: Based on the uniformity evaluation score and feature vector set, a classification algorithm is used to classify the test screen of the display module, and the corresponding abnormal category is obtained from the classification result; S5: Based on the anomaly category, brightness gradient characteristics, and color deviation characteristics, adjust the position parameters of the display module, and re-acquire the test image of the display module to be tested through the acquisition device; S6: Obtain brightness and contrast characteristics from the test screen of the display module to be tested, calculate the corresponding brightness uniformity value and contrast uniformity value, and use a multi-parameter fusion processing method to jointly calculate the brightness uniformity value and contrast uniformity value to obtain the comprehensive performance value. S7: Determine whether the comprehensive performance value meets the preset standard, output the judgment result, and generate the final test report by combining the comprehensive performance value and the judgment result.

[0019] In the automated testing method for the optical performance of display modules, brightness gradient and color deviation features are extracted from the acquired images to construct a feature vector set. The feature amplitude is then statistically analyzed to determine if there is an acquisition angle deviation. If a deviation is found, region segmentation is performed to accurately assess image uniformity. A classification algorithm is then used to identify the anomaly category. Based on this, the display module's position parameters are adjusted, the test image is re-acquired, and the brightness and contrast uniformity values ​​are fused to calculate the overall performance value. Otherwise, the brightness and contrast uniformity values ​​of the currently acquired image are directly fused to calculate the overall performance value, ultimately generating a test report. This solution effectively identifies the presence of acquisition angle deviation anomalies through multi-parameter fusion and dynamic adjustment strategies. By adjusting the display module's position according to the anomaly type and re-acquiring, the accuracy and reliability of display module testing are improved.

[0020] It is worth noting that in step S2, the gradient Gx in the horizontal direction and the gradient Gy in the vertical direction of each pixel are calculated based on the Sobel operator on the acquired image, where Gx represents the intensity of the horizontal brightness change and Gy represents the intensity of the vertical brightness change; the gradient magnitude M is obtained from Gx and Gy. ; By statistically analyzing the gradient magnitude M of each pixel in the entire acquired image, a set of brightness gradient features is obtained. The brightness gradient feature set is subjected to min-max normalization, and all gradient magnitudes M are linearly scaled to between 0 and 1 to obtain the brightness gradient features.

[0021] In one embodiment, after acquiring the test image of the display module through a data acquisition device, edge detection processing is performed on the acquired image using the Sobel operator. For example, when calculating the gradient of a pixel, the Sobel operator filters the image using convolution kernels. The horizontal kernel is such as [-1,0,1;-2,0,2;-1,0,1], and the vertical kernel is similar but transposed, such as [-1,-2,-1;0,0,0;1,2,1]. For each pixel, Gx is obtained by weighted summation of the horizontal kernel and the brightness values ​​of neighboring pixels, and Gy is obtained by weighted summation of the vertical kernel and the brightness values ​​of neighboring pixels. Thus, Gx reflects the intensity change of horizontal edges in the image, and Gy captures the response intensity of vertical edges. Subsequently, the gradient magnitude M is calculated from Gx and Gy, obtained by taking the square root of the sum of squares. This magnitude represents the brightness intensity of the pixel.

[0022] Next, the M values ​​of all pixels are counted to form a set of brightness gradient features, and then the maximum and minimum normalization is performed. That is, the maximum value M_max and the minimum value M_min in the set of brightness gradient features are found, and then each M is normalized to (M-M_min) / (M_max-M_min). In this way, all values ​​are mapped to the interval between 0 and 1, and the brightness gradient features of each pixel are obtained.

[0023] Preferably, in step S2, the acquired image is converted from the RGB color space to the Lab color space to obtain a Lab image; In Lab images, standard whiteboard data is used as the standard color reference. In the standard whiteboard data, brightness... The color axis from green to red The color axis from blue to yellow The Lab value of each pixel and the Euclidean distance d from the standard color reference are calculated as the color difference intensity, where The Lab value for each pixel includes brightness. The color axis from green to red and the color axis from blue to yellow ; The pixel color difference calculation results are subjected to maximum and minimum normalization processing, and all color difference intensities d are linearly scaled to between 0 and 1 to obtain color deviation characteristics; The brightness gradient feature and color deviation feature of each pixel are sequentially concatenated into a two-dimensional feature. Then, all two-dimensional features are arranged according to the spatial arrangement of the pixels to encapsulate a set of feature vectors.

[0024] In color processing, the RGB image is converted to the Lab color space using the CIE Lab formula, where the L channel represents lightness, the a channel represents the green-to-red color axis, and the b channel represents the blue-to-yellow color axis. Using a standard white board as a reference, the Euclidean distance d of each pixel is calculated, derived by taking the square root of the sum of the squares of the differences between the channels. For example, in display module testing, if the Lab value of a pixel is (95, 2, 3), then... It is 95. It is 2. If the value is 3, then the calculated value of d is approximately 6.16, representing the intensity of the deviation from the ideal white. Similarly, all d values ​​are subjected to max-min normalization and linearly scaled to 0 to 1.

[0025] Finally, the brightness gradient features and color deviation features of each pixel are concatenated into a two-dimensional vector and arranged into a set according to the spatial order of the image. For example, in a 1920x1080 image, a set containing about 2 million two-dimensional vectors is formed for subsequent analysis.

[0026] Optionally, in step S3, the feature magnitude z of each two-dimensional feature is calculated using the modulus formula, where the modulus formula is: , where x represents the brightness gradient feature and y represents the color deviation feature, thus obtaining the amplitude set; The number of two-dimensional features whose feature amplitude z exceeds a preset deviation value is counted from the amplitude set. If the number exceeds a preset threshold, the U-Net model is applied to the acquired image for region segmentation. The U-Net model takes the acquired image as input and outputs a segmentation mask. The list of boundary coordinates is extracted from the segmentation mask using OpenCV's findContours function to obtain a set of regions; For each region in the region set, the brightness gradient features of all pixels in the region are extracted, summed, and then divided by the number of pixels to obtain the average brightness of the region. Then, the brightness variance value of each region is calculated using the variance formula, which is the sum of the squares of the brightness gradient features minus the average brightness of the region divided by the number of pixels in the region. The brightness variance value represents the difference in brightness distribution. The uniformity assessment score is obtained by summing all the brightness variance values ​​and dividing by the number of regions.

[0027] In one embodiment, the feature magnitude z of each two-dimensional feature is calculated using a modulus formula. For example, assuming a two-dimensional vector is (0.5, 0.3), the magnitude z is... This yields the amplitude set of all two-dimensional features.

[0028] For example, the number of two-dimensional features whose amplitude z exceeds a preset deviation value is counted from the amplitude set. The preset deviation value is obtained by summing the amplitude values ​​of historical image data and dividing by the number of data. Specifically, in actual business processes, amplitude data of multiple historical test images are first collected. For example, if the total amplitude sum of 10 images is 500 and the number of data is 1000, then the preset deviation value is 500 / 1000=0.5. If the number of z values ​​exceeding 0.5 in the current amplitude set is 200 and the preset threshold is 150, then the number exceeds the threshold. Therefore, the U-Net model is applied to the acquired images for region segmentation processing.

[0029] It should be noted that the U-Net model is a semantic segmentation architecture based on convolutional neural networks. Its structure is U-shaped, including a downsampling path for capturing contextual information and an upsampling path for precise localization. The input is a captured image, such as a display screen image in RGB format, and the output is a segmentation mask. This mask marks different regions at the pixel level. For example, the model learns to distinguish between normal display areas and potential abnormal areas through training, thereby generating a binary mask, where white represents potential abnormal areas and black represents normal display areas, which facilitates subsequent extraction.

[0030] In one embodiment, the list of boundary coordinates is extracted from the segmentation mask using OpenCV's findContours function to obtain a set of regions. For example, the findContours function detects continuous boundary points by traversing the mask image and outputs the contour hierarchy and a list of coordinate points. In the display module test, for a mask containing potentially abnormal regions, this function can extract the coordinate sequence of multiple closed contours to form a set of regions.

[0031] For example, considering one region in a set of regions, this region has 4 pixels with brightness gradient features of 0.1, 0.2, 0.3, and 0.4, respectively. The average brightness is (0.1+0.2+0.3+0.4) / 4=0.25, and the brightness variance is... This represents the difference in brightness distribution. In one embodiment, the average variance is calculated by summing all brightness variance values ​​and dividing by the number of regions. This average variance is used as a uniformity assessment score for subsequent evaluation.

[0032] Specifically, in step S4, multiple sets of display module test data are obtained from the historical database. Each set of data includes labels for abnormal categories, uniformity evaluation scores, brightness gradient features of each pixel, and color deviation features of each pixel. Based on the test data, the uniformity evaluation score is used as the first feature; The average value of the brightness gradient features of all pixels is calculated to obtain the second feature; The average value of the color deviation features of all pixels is calculated to obtain the third feature; The first feature, the second feature, and the third feature are combined into a joint input vector; The label of the anomaly category is used as the target output, which together with the joint input vector constitutes a classification training sample set; Using the aforementioned classification training sample set, a multi-class support vector machine model is trained with a one-to-one residual strategy. The model outputs the confidence score of the joint input feature vector belonging to each preset anomaly category.

[0033] In one embodiment, multiple sets of display module test data are obtained from a historical database. These data cover labels for all anomaly categories, with each anomaly category represented by a corresponding numerical code as a label. The test data also includes uniformity evaluation scores, brightness gradient features of each pixel, and color deviation features, thereby providing a foundation for subsequent model training.

[0034] In display module detection, the uniformity evaluation score is first used as the first feature based on the test data, reflecting the stability of the overall brightness distribution of the image. Then, the average value of the brightness gradient features of all pixels is calculated as the second feature. Assuming an image has 10,000 pixels and the sum of the brightness gradient features is 50,000, the average value of the brightness gradient features is 5, thus capturing global brightness fluctuations and effectively reducing the dimensionality of the subsequently formed joint input vector. Similarly, the average value of the color deviation features is calculated as the third feature. For example, if the sum of the color deviation features is 30,000 and the number of pixels is 10,000, the average value of the color deviation features is 3, thus capturing global color fluctuations and effectively reducing the dimensionality of the subsequently formed joint input vector. These features are combined into a joint input vector, which, together with the label of the anomaly category, forms a classification training sample set.

[0035] Building upon this, a one-to-one residual strategy is employed to train a multi-class support vector machine (SVM) model. SVM is a supervised learning algorithm that separates different classes by finding the maximum margin hyperplane. The one-to-one residual strategy decomposes the multi-class problem into multiple binary sub-problems. For example, for three anomaly classes with labels 1, 2, and 3, classifiers Avs1, Avs2, and Bvs3 are constructed, and the final class is determined through a voting mechanism. The model outputs a confidence score for the joint input vector belonging to each preset anomaly class. The confidence score is calculated based on a decision function, such as the signed distance from the hyperplane, which represents the support. This optimizes the classification boundary during training to improve accuracy.

[0036] It is worth noting that in step S4, a uniformity evaluation score for the acquired image is obtained. Calculate the average value of the brightness gradient features of all pixels in the feature vector set of the acquired image; Calculate the average value of the color deviation features of all pixels in the feature vector set of the acquired image; The average value of the uniformity evaluation score, the average value of the brightness gradient feature, and the average value of the color deviation feature of the acquired image are used to construct the current input vector; Input the current input vector into the trained multi-class support vector machine model; The multi-class support vector machine model uses a decision function method to calculate the confidence score of the current input vector for each anomaly category; The anomaly category corresponding to the highest confidence score is obtained as the output category of the acquired image.

[0037] Beneficial effects: Specifically, this method improves classification accuracy through feature integration.

[0038] Preferably, in step S5, the average value of the brightness gradient feature and the average value of the color deviation feature are obtained and combined to form a normalized feature vector. For the labels of the aforementioned anomaly categories, the target vector of the standard module is determined by a lookup table. The target vector includes the corresponding standard values ​​of brightness gradient features and color deviation features. Based on the normalized feature vector and the target vector, the Euclidean distance between them is calculated as the fitness function; The position of the population is initialized using a differential evolution algorithm, where the position represents the orientation of the display module; Each position is evaluated according to the fitness function, and the optimal position is iteratively updated to minimize the fitness function, thus obtaining the final orientation of the display module. Based on the final orientation and the current orientation of the display module, optimized test configuration data is generated; The test configuration data is written into the driver chip of the display module, and the angle of the display module is adjusted to complete the orientation alignment correction. The test image displayed by the display module is obtained by acquiring the test screen through the acquisition device, and a resampled image is obtained.

[0039] In one embodiment, the average value of the brightness gradient feature and the average value of the color deviation feature in the test screen of the display module are first obtained and combined into a normalized feature vector, thereby providing basic data for subsequent comparisons and reducing the dimensionality of data processing.

[0040] Next, the Euclidean distance is calculated as the fitness function based on the normalized feature vector and the target vector. Specifically, the Euclidean distance is obtained by summing the squared differences of the components of the two vectors and then taking the square root. For example, if the normalized feature vector is (0.6, 0.3) and the target vector is (0.5, 0.2), the Euclidean distance is 0.1414. This distance quantifies the deviation of the current module from the standard and is used for optimization evaluation. In this embodiment, for different anomaly categories, there are corresponding target vectors for different standard modules, and the target vectors of these standard modules and the labels of the anomaly categories are in a one-to-one correspondence in the mapping table.

[0041] In one embodiment, a differential evolution algorithm is used to initialize the positions of the population, which represent the possible orientation angles of the display module. For example, 10 individuals are randomly generated, and the position of each individual is represented by Euler angles. The algorithm iteratively updates the population through mutation, crossover, and selection operations. For example, during the mutation phase, three individuals are randomly selected to generate a differential vector, which is added to the target individual. During crossover, components are swapped with a probability of 0.5. During selection, the fitness function value is compared and the smaller value is retained. After multiple iterations, such as 50 generations, the fitness is minimized to obtain the optimal orientation. Specifically, a linear interpolation method is used to pre-establish the relationship between the position (represented by Euler angles α, β, and γ) and the average values ​​of the brightness gradient features and color deviation features in the normalized feature vector. After iteratively updating the position, the average values ​​of the corresponding brightness gradient features and color deviation features can be obtained based on the Euler angles of the position through the pre-established relationship. In the pre-establishment process, the position orientation (Euler angles) of the display module can be collected first, and then the brightness gradient features and color deviation features of each pixel can be obtained according to steps S1 and S2. The average value of the brightness gradient features is calculated based on the brightness gradient features, and the average value of the color deviation features is calculated based on the color deviation features. This yields a data set that reflects a position and includes the Euler angles, the average values ​​of the brightness gradient features, and the average values ​​of the color deviation features. Then, the Euler angles are used as the independent variable, and the average values ​​of the brightness gradient features and the average values ​​of the color deviation features are used as the dependent variables to perform linear interpolation.

[0042] Then, based on this final orientation and the current orientation, optimized test configuration data is generated. For example, the difference is calculated to generate an angle adjustment command, which is then used as the optimized test configuration data. The angle adjustment command is then written to the driver chip, and the display module angle is adjusted by updating the chip's internal registers to complete the alignment correction. Finally, the adjusted test image is acquired using a data acquisition device such as a high-definition camera, and a resampled image is obtained for further verification.

[0043] Optionally, in step S6, the test screen of the display module to be tested is converted into a grayscale image, the grayscale value of each pixel is calculated for the grayscale image to obtain a brightness matrix, the standard deviation and the average value of the brightness matrix are calculated, and the standard deviation is divided by the average value to obtain the brightness uniformity value. Based on the brightness matrix, calculate the average difference between each pixel and all its neighboring pixels to obtain the contrast matrix; calculate the standard deviation and the average of the contrast matrix, and divide the standard deviation by the average to obtain the contrast uniformity value. Using preset weighting coefficients, the brightness uniformity value and the contrast uniformity value are weighted and summed to obtain the overall performance value.

[0044] In one embodiment, the test screen of the display module to be tested is first converted into a grayscale image. For example, the RGB image is weighted and averaged using an image processing tool, where the weights of red, green and blue are 0.3, 0.59 and 0.11, respectively, to obtain the grayscale value of each pixel and form an overall grayscale image for subsequent analysis.

[0045] For the obtained grayscale image, the grayscale value of each pixel is calculated to construct a brightness matrix. Specifically, this process involves traversing all pixels of the image. For example, assuming the image resolution is 1920x1080, a matrix of corresponding size is generated, where each element corresponds to the grayscale value of a pixel. For instance, the grayscale value at position (0,0) in the matrix is ​​128, representing the brightness level of that point. This matrix representation allows for further quantification of the brightness distribution of the display module. Next, the standard deviation and mean of the brightness matrix are calculated using the NumPy library. For example, the mean of all elements is first calculated, such as 150, and then the standard deviation is calculated, such as 20. Dividing the standard deviation by the mean yields the brightness uniformity value, such as 0.133, which reflects the degree of consistency in brightness.

[0046] Furthermore, for the construction of the contrast matrix, the Sobel operator is used to calculate the average of the differences between each pixel and all its neighboring pixels. For example, for the center pixel, the Sobel operator applies horizontal and vertical convolution kernels, and the absolute values ​​are averaged to form the contrast matrix, where each element represents the local contrast intensity. Similarly, NumPy is used to calculate its standard deviation and mean, and the contrast uniformity value is obtained.

[0047] Finally, using preset weighting coefficients, the above uniformity values ​​are weighted and summed to obtain the comprehensive performance value.

[0048] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. An automated testing method for the optical performance of a display module, characterized in that, Includes the following steps: S1: Acquire the test screen of the display module to be tested as the acquired image; S2: Extract brightness gradient features and color deviation features from the acquired images to determine the feature vector set; S3: Calculate the feature amplitude of each two-dimensional feature in the feature vector set, and count the number of feature amplitudes that exceed the preset deviation value; if the number exceeds the preset threshold, perform region segmentation processing on the acquired image, calculate the brightness uniformity index of the segmented region, determine the uniformity evaluation score, and then execute step S4; otherwise, execute step S6. S4: Based on the uniformity evaluation score and feature vector set, a classification algorithm is used to classify the test screen of the display module, and the corresponding abnormal category is obtained from the classification result; S5: Based on the anomaly category, brightness gradient characteristics, and color deviation characteristics, adjust the position parameters of the display module, and re-acquire the test image of the display module to be tested through the acquisition device; S6: Obtain brightness and contrast characteristics from the test screen of the display module to be tested, calculate the corresponding brightness uniformity value and contrast uniformity value, and use a multi-parameter fusion processing method to jointly calculate the brightness uniformity value and contrast uniformity value to obtain the comprehensive performance value. S7: Determine whether the comprehensive performance value meets the preset standard, output the judgment result, and generate the final test report by combining the comprehensive performance value and the judgment result.

2. The automated testing method for the optical performance of a display module according to claim 1, characterized in that: In step S2, the gradient Gx in the horizontal direction and the gradient Gy in the vertical direction of each pixel are calculated on the acquired image based on the Sobel operator, where Gx represents the intensity of the horizontal brightness change and Gy represents the intensity of the vertical brightness change; the gradient magnitude M is obtained from Gx and Gy. ; By statistically analyzing the gradient magnitude M of each pixel in the entire acquired image, a set of brightness gradient features is obtained. The brightness gradient feature set is subjected to min-max normalization, and all gradient magnitudes M are linearly scaled to between 0 and 1 to obtain the brightness gradient features.

3. The automated testing method for the optical performance of a display module according to claim 1, characterized in that: In step S2, the acquired image is converted from the RGB color space to the Lab color space to obtain a Lab image; In the Lab image, standard whiteboard data is used as the standard color reference, and the Euclidean distance d between the Lab value of each pixel and the standard color reference is calculated as the color difference intensity. The pixel color difference calculation results are subjected to maximum and minimum normalization processing, and all color difference intensities d are linearly scaled to between 0 and 1 to obtain color deviation characteristics; The brightness gradient feature and color deviation feature of each pixel are sequentially concatenated into a two-dimensional feature. Then, all two-dimensional features are arranged according to the spatial arrangement of the pixels to encapsulate a set of feature vectors.

4. The automated testing method for the optical performance of a display module according to claim 1, characterized in that: In step S3, the feature magnitude z of each two-dimensional feature is calculated using the modulus formula, which is: , where x represents the brightness gradient feature and y represents the color deviation feature, thus obtaining the amplitude set; The number of two-dimensional features whose feature amplitude z exceeds a preset deviation value is counted from the amplitude set. If the number exceeds a preset threshold, the U-Net model is applied to the acquired image for region segmentation. The U-Net model takes the acquired image as input and outputs a segmentation mask. The list of boundary coordinates is extracted from the segmentation mask using the findContours function, resulting in a set of regions; For each region in the region set, the brightness gradient features of all pixels in the region are extracted, summed, and then divided by the number of pixels to obtain the average brightness of the region. Then, the brightness variance value of each region is calculated. The uniformity assessment score is obtained by summing all the brightness variance values ​​and dividing by the number of regions.

5. The automated testing method for the optical performance of a display module according to claim 1, characterized in that: In step S4, multiple sets of display module test data are obtained from the historical database. Each set of data includes labels for anomaly categories, uniformity evaluation scores, brightness gradient features of each pixel, and color deviation features of each pixel. Based on the test data, the uniformity evaluation score is used as the first feature; The average value of the brightness gradient features of all pixels is calculated to obtain the second feature; The average value of the color deviation features of all pixels is calculated to obtain the third feature; The first feature, the second feature, and the third feature are combined into a joint input vector; The label of the anomaly category is used as the target output, which together with the joint input vector constitutes a classification training sample set; Using the aforementioned classification training sample set, a multi-class support vector machine model is trained with a one-to-one residual strategy. The model outputs the confidence score of the joint input feature vector belonging to each preset anomaly category.

6. The automated testing method for the optical performance of a display module according to claim 5, characterized in that: In step S4, a uniformity evaluation score for the acquired image is obtained. Calculate the average value of the brightness gradient features of all pixels in the feature vector set of the acquired image; Calculate the average value of the color deviation features of all pixels in the feature vector set of the acquired image; The average value of the uniformity evaluation score, the average value of the brightness gradient feature, and the average value of the color deviation feature of the acquired image are used to construct the current input vector; Input the current input vector into the trained multi-class support vector machine model; The multi-class support vector machine model uses a decision function method to calculate the confidence score of the current input vector for each anomaly category; The anomaly category corresponding to the highest confidence score is obtained as the output category of the acquired image.

7. The automated testing method for the optical performance of a display module according to claim 1, characterized in that: In step S5, the average value of the brightness gradient feature and the average value of the color deviation feature are obtained and combined to form a normalized feature vector. For the labels of the aforementioned anomaly categories, the target vector of the standard module is determined by a lookup table. The target vector includes the corresponding standard values ​​of brightness gradient features and color deviation features. Based on the normalized feature vector and the target vector, the Euclidean distance between them is calculated as the fitness function; The position of the population is initialized using a differential evolution algorithm, where the position represents the orientation of the display module; Each position is evaluated according to the fitness function, and the optimal position is iteratively updated to minimize the fitness function, thus obtaining the final orientation of the display module. Based on the final orientation and the current orientation of the display module, optimized test configuration data is generated; The test configuration data is written into the driver chip of the display module, and the angle of the display module is adjusted to complete the orientation alignment correction. The test image displayed by the display module is obtained by acquiring the test screen through the acquisition device, and a resampled image is obtained.

8. The automated testing method for the optical performance of a display module according to claim 1, characterized in that: In step S6, the test screen of the display module to be tested is converted into a grayscale image, and the grayscale value of each pixel is calculated for the grayscale image to obtain the brightness matrix; Calculate the standard deviation and mean of the luminance matrix, and divide the standard deviation by the mean to obtain the luminance uniformity value; Based on the brightness matrix, the average difference between each pixel and all its neighboring pixels is calculated to obtain the contrast matrix; Calculate the standard deviation and mean of the contrast matrix, and divide the standard deviation by the mean to obtain the contrast uniformity value; Using preset weighting coefficients, the brightness uniformity value and the contrast uniformity value are weighted and summed to obtain the overall performance value.

Images