A performance detection method for display screen

Abstract

The present invention relates to the field of display technology, and in particular to a performance detection method for a display screen. The method comprises the following steps: obtaining a historical display screen bright screen display image and performing defect detection to obtain a historical display screen bright screen defect annotated image; constructing a Mura defect detection model based on the historical display screen bright screen defect annotated image; obtaining a display image set of a display screen to be detected and performing performance detection to obtain first display screen performance detection data; extracting an off-screen display image from the display image set of the display screen to be detected and performing performance detection to obtain second display screen performance detection data; calculating the first display screen performance detection data and the second display screen performance detection data to obtain a performance evaluation coefficient of the display screen to be detected, and performing data merging to obtain performance evaluation data of the display screen to be detected; and constructing a display screen digital twin model based on the performance evaluation data of the display screen to be detected. The present invention performs performance detection on a display screen based on data mining.

Description

Technical Field

[0001] The present invention relates to the field of display technology, and in particular to a performance detection method for a display screen. Background Art

[0002] In modern technology, display screens have been widely used in various electronic devices, such as TVs, mobile phones, computers, etc. To ensure that the performance of the display screen reaches the expected level, it is crucial to detect the performance characteristics of the display screen. Traditional display screen detection methods are usually based on human eye observation and subjective evaluation, which are highly subjective, time-consuming and labor-intensive, and cannot provide accurate quantitative data. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a performance detection method for a display screen to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for detecting the performance of a display screen comprises the following steps:

[0005] Step S1: obtaining a historical display screen bright screen display image, performing defect detection on the historical display screen bright screen display image, thereby obtaining a historical display screen bright screen defect annotated image;

[0006] Step S2: constructing a model for the historical display screen bright screen defect annotated images based on a deep learning method, thereby obtaining a Mura defect detection model;

[0007] Step S3: obtaining a display image set of the display screen to be inspected, and performing a first display screen performance test on the display image set of the display screen to be inspected using a Mura defect detection model, thereby obtaining first display screen performance test data;

[0008] Step S4: extracting the off-screen display image from the display image set to be detected, thereby obtaining the off-screen display image of the display screen, and performing a second display screen performance test on the off-screen display image of the display screen, thereby obtaining second display screen performance test data;

[0009] Step S5: calculating a comprehensive evaluation coefficient for the first display screen performance test data and the second display screen performance test data, thereby obtaining a performance evaluation coefficient for the display screen to be tested, and merging the first display screen performance test data and the second display screen performance test data according to the performance evaluation coefficient for the display screen to be tested, thereby obtaining performance evaluation data for the display screen to be tested;

[0010] Step S6: construct a digital twin model of the display screen according to the performance evaluation data of the display screen to be tested, and upload the digital twin model of the display screen to the visualization platform to perform the 3D visualization task of the display screen defects.

[0011] The present invention can obtain a historical display screen bright screen defect annotated image by acquiring a historical display screen bright screen display image and performing defect detection. Defect sample data is established, which contains various defects that may occur when the display screen is bright. These data will be used for the construction and training of the deep learning model in the subsequent process. The historical display screen bright screen defect annotated image is modeled based on the deep learning method to obtain a Mura defect detection model. A large number of bright screen defect annotated images are used, and the model is constructed by a deep learning algorithm to improve the accuracy and consistency of the detection and reduce the subjectivity. Through the deep learning model, various defects in the bright screen display image can be accurately detected, such as missing pixels, uneven brightness or abnormal colors. The Mura defect detection model can be used to detect defects in the display screen to be detected in the subsequent process. By using the Mura defect detection model to perform a first display screen performance test on the display image set of the display screen to be detected, the first display screen performance test data can be obtained. It is possible to promptly discover defects that may exist when the display screen to be detected is bright, and obtain quantitative data on the performance of the display screen. These data can help evaluate and judge the quality of the display screen. The display screen to be detected is subjected to off-screen display image extraction, and the off-screen display image is subjected to a second display screen performance test, thereby obtaining the second display screen performance test data. This can detect possible problems of the display screen to be tested in the off state, such as bad pixels, dark spots or uneven brightness. The performance test data of the second display screen can further improve the evaluation of the performance of the display screen. The display screen is tested in an automated manner, which greatly reduces the manpower and time costs and improves efficiency. The performance evaluation coefficient of the display screen to be tested can be obtained by calculating the comprehensive evaluation coefficient of the performance test data of the first display screen and the performance test data of the second display screen. Based on objective data calculation, the influence of subjective evaluation is reduced, and a more accurate performance evaluation result is provided. The data obtained from the two tests are comprehensively analyzed and weighed to obtain the evaluation result of the overall performance of the display screen. By calculating the evaluation coefficient, the performance status of the display screen to be tested can be more accurately evaluated. A digital twin model of the display screen is constructed according to the performance evaluation data of the display screen to be tested, and it is uploaded to the visualization platform to perform the 3D visualization task of the display screen defects. The performance data and defect information of the display screen are combined with the digital twin model to realize the visualization of the display screen defects. In this way, the defect problems of the display screen can be intuitively observed and analyzed, and corresponding measures can be taken to repair or improve them. Through the visual presentation of the digital model, the defect problems of the display screen are more intuitive and clear, reducing the deviation of individual subjective evaluation.

[0012] Optionally, step S1 specifically includes:

[0013] Step S11: Obtain the historical display screen bright screen display image;

[0014] Step S12: converting the historical display screen bright screen display image into a grayscale image, thereby obtaining a display screen bright screen display grayscale image;

[0015] Step S13: performing background segmentation on the grayscale image displayed on the bright screen of the display screen by using a display screen image segmentation algorithm, thereby obtaining a display foreground image;

[0016] Step S14: binarizing the display foreground image to obtain a display binary image;

[0017] Step S15: performing connected domain defect determination on the displayed binary image, thereby obtaining a historical display screen bright screen defect annotated image.

[0018] The present invention obtains the bright screen display images of historical display screens, and collects and obtains a large number of bright screen display images of historical display screens. These images are the basic data required for subsequent detection and evaluation. By collecting a large number of images, a rich data set can be established, which is helpful for model training and performance improvement. The bright screen display images of the historical display screens are converted into grayscale images, which simplifies the complexity of subsequent processing while retaining the key information of the image. The grayscale image only contains brightness information, without considering the color. This can reduce the resource consumption of calculation and storage, and is more convenient for image processing and analysis. The bright screen display grayscale image of the display screen is segmented by the display screen image segmentation algorithm, and the display foreground image can be obtained by segmentation, and the interference of the background is removed, so that the subsequent processing is more focused on the display content, and the accuracy and reliability of defect detection are improved. The binarization process converts the image into a binary image containing only black and white colors. The image can be further simplified, making the subsequent defect detection simpler and more efficient. The white in the binary image represents the display content, and the black represents the non-display content, so that the edges and details of the defects are clearer and more prominent. Conduct connected domain defect determination on binary images; connected domain defect determination refers to determining the defective parts by analyzing and determining the connected areas in binary images. Locate and annotate defective areas in historical display screens to provide annotation data for subsequent model training and detection. These annotated images can be used to train deep learning models to improve the accuracy and robustness of defect detection.

[0019] Optionally, step S13 is specifically:

[0020] Step S131: extracting grayscale pixel values ​​of the grayscale image displayed on the bright screen of the display screen, thereby obtaining the grayscale pixel values ​​of the bright screen of the display screen;

[0021] Step S132: performing statistical analysis on the grayscale pixel values ​​of the display screen to obtain a grayscale pixel value threshold;

[0022] Step S133: performing edge segmentation on the grayscale image displayed on the display screen by using a display screen image segmentation algorithm according to the grayscale pixel value threshold, thereby obtaining a display screen edge segmentation grayscale image;

[0023] Step S134: Gaussian blurring the segmented grayscale image at the edge of the display screen to obtain an image to be background-eliminated;

[0024] Step S135: performing background removal on the image to be background-removed using a background removal method based on the domain information, thereby obtaining a display foreground image.

[0025] The present invention extracts grayscale pixel values ​​from a grayscale image displayed on a bright screen of a display screen; converts a color image into a grayscale image and extracts grayscale pixel values. Grayscale pixel values ​​reflect the brightness information of each pixel in an image. By extracting grayscale pixel values, subsequent statistical analysis and processing can be performed more conveniently. Statistical analysis is performed on the grayscale pixel values ​​of the bright screen of the display screen; the beneficial effect of statistically analyzing the grayscale pixel values ​​is to determine a suitable grayscale pixel value threshold value according to the brightness distribution characteristics of the image. The selection of the threshold value is based on the brightness difference between the foreground and the background in the image, which will help the subsequent image segmentation to accurately distinguish the display foreground from the background. The edge segmentation of the grayscale image displayed on the bright screen of the display screen is performed by a display screen image segmentation algorithm according to the grayscale pixel value threshold value; image segmentation is performed based on the grayscale pixel value threshold value, and the image is divided into the foreground and the background. Through edge segmentation, the edge information of the display screen can be obtained, which is very helpful for further processing and analysis of the display foreground and background areas. Gaussian blur is performed on the grayscale image segmented by the edge of the display screen; Gaussian blur is a commonly used image processing operation that can reduce noise and interference in the image and smooth the detailed information of the image. By performing Gaussian blur on the edge segmentation grayscale image, the edge of the display foreground can be made smoother and more accurate image data can be provided for subsequent background removal. The background removal method based on domain information removes the background of the background-removed image; the background removal method based on domain information removes the background part of the image by processing the background-removed image. Background removal extracts the foreground part of the display screen, so that subsequent defect detection and analysis can focus on the display content, improving the accuracy and reliability of detection.

[0026] Optionally, the function formula of the display screen image segmentation algorithm in step S133 is specifically:

[0027]

[0028] Where E is the edge intensity value, I is the gray value of the pixel, S is the gray pixel value threshold, k is the edge enhancement coefficient, α is the image enhancement index, β is the translation parameter, and γ is the exponential parameter. is the gradient operator, x is the horizontal coordinate of the pixel in the image, y is the vertical coordinate of the pixel in the image, and e is the base of the natural logarithm.

[0029] The present invention constructs a function formula of a display screen image segmentation algorithm, which is used to calculate the edge intensity value of a grayscale image displayed on a bright screen of a display screen. The formula fully considers the grayscale value I of the pixel that affects the edge intensity value E, the grayscale pixel value threshold S, the edge enhancement coefficient k, the image enhancement index α, the translation parameter β, the exponential parameter γ, the gradient operator The horizontal coordinate x of the pixel in the image, the vertical coordinate y of the pixel in the image, and the base e of the natural logarithm form a functional relationship:

[0030]

[0031] in Represents the gradient operator, which is used to calculate the gradient of the image. Application Compute gradient information. represents the Euclidean norm, which calculates the magnitude of the gradient. Where x and y represent the gradient values ​​in the corresponding directions, and β is a translation parameter. By adding β inside the square root, the curve is translated to adapt to the gradient value distribution of different images. γ represents the exponential parameter, which is used to adjust By adjusting γ, the response of edge enhancement to the gradient value can be controlled. I is the gray value of the pixel, which represents the input gray image. S represents the gray pixel value threshold, which is used to divide the gray image into two parts: bright and dark. The edge will be generated in the area where the brightness is higher than the threshold. k represents the edge enhancement coefficient, which is used to adjust the strength of the edge. Increasing k will make the edge more obvious. α represents the image enhancement index, which is used to adjust the effect of brightness on the edge. Increasing α will enhance the contribution of brightness to the edge. Used to adjust the impact of edge enhancement. The magnitude of this function decreases as the brightness of the input grayscale image increases. is a logic function used to limit the brightness range of edge generation. Larger γ and (SI) will reduce the brightness of edge generation. represents the calculation of gradient, and generates edge image through gradient operation. Then, the edge image is Multiply to adjust the edge strength and brightness. Finally, multiply by Used to control the brightness range of edge generation. The gradient operation and edge strength adjustment parts in the formula can effectively extract the edge information of the image. By calculating the gradient and adjusting the edge strength, the edge of the image can be separated from the background, thereby realizing applications such as image segmentation and target detection. The edge enhancement part in the formula can enhance the edges in the image, making them more obvious and clear. By adjusting the edge enhancement coefficient and the image enhancement index, the edge strength and the degree of response to brightness can be controlled, thereby improving the visual effect and recognition of the image. The formula introduces multiple parameters, such as the translation parameter β, the exponential parameter γ, and the grayscale pixel value threshold S. These parameters can be adjusted according to the specific application and image characteristics to adapt to different image scenes and needs. By adjusting these parameters, the performance and applicability of the algorithm can be optimized to achieve better image segmentation effects. By introducing logical function terms in the formula The brightness range of edge generation can be limited to make the edges of darker areas more obvious. This can increase the contrast of the image and improve visual perception.

[0032] Optionally, step S15 is specifically:

[0033] Denoising the binary image based on morphological filtering to obtain a denoised binary image.

[0034] Calculate the connected domain area of ​​the denoised binary image to obtain the connected domain area data;

[0035] The spectrum data of the connected domain area is converted into spectrum data by using a one-dimensional discrete Fourier transform method, so as to obtain the spectrum data of the connected domain area;

[0036] Perform abnormal band detection on the connected domain area spectrum data to obtain the abnormal band of the connected domain area;

[0037] The image of the abnormal band of the connected domain area is converted by the inverse Fourier transform method to obtain the bright screen defect image;

[0038] Defects of the denoised binary image are annotated according to the bright screen defect image, so as to obtain a bright screen defect annotated image of a historical display screen.

[0039] The present invention performs denoising on a binary image based on a denoising method of morphological filtering; morphological filtering is an effective image denoising method, which can remove noise in an image and retain the main target structure by dilating and corroding structural elements in an image. The application of the morphological filtering method to denoise a binary image can eliminate noise points and small discontinuous areas in the image, thereby obtaining a clear and continuous denoised binary image. The connected domain area is calculated for the denoised binary image; after the denoised binary image is obtained, the size information of each connected domain can be obtained by calculating the area of ​​the connected domain, that is, the size of the adjacent pixel set. This helps to count the size distribution of different connected domains in the image and obtain meaningful image attributes therefrom. The connected domain area data is converted into spectrum data by a one-dimensional discrete Fourier transform method; the connected domain area data can be converted into spectrum data by performing DFT processing on the connected domain area data. Such spectrum data can better reveal the spectrum characteristics of the connected domain area distribution, such as frequency components, intensity and other information. The connected domain area spectrum data is detected by abnormal bands; the connected domain area spectrum data is detected by abnormal bands, and abnormal bands in the spectrum can be identified. These abnormal bands may correspond to special structures or noise in the image. By detecting and analyzing these abnormal bands, connected domains with special properties or abnormal characteristics can be extracted. The abnormal bands of the connected domain area are converted into images by the inverse Fourier transform method; the spectrum data extracted from the abnormal bands can be converted back to the spatial domain image by applying the inverse Fourier transform. In this way, a bright screen defect image can be obtained, which contains areas with abnormal brightness, which is helpful for further analysis and processing. Defects are annotated on the denoised binary image based on the bright screen defect image; based on the bright screen defect image, the corresponding defect area can be annotated to visualize and locate it in the image. This helps to quickly identify and analyze bright screen defects on historical display screens, and perform subsequent processing and repair.

[0040] Optionally, step S2 specifically includes:

[0041] Performing image enhancement on the historical display screen bright screen defect annotated image, thereby obtaining an enhanced historical display screen bright screen defect annotated image;

[0042] The enhanced historical display screen bright screen defect annotation image is converted into a label format, thereby obtaining an enhanced historical display screen bright screen defect annotation label;

[0043] The enhanced historical display screen bright screen defect annotation labels are divided into a modeling training set and a modeling test set according to a preset division ratio;

[0044] Model the modeling training set and the modeling test set based on the YOLO3 algorithm to obtain a defect detection model;

[0045] The Gaussian error evaluation iteration is performed on the defect detection model according to the modeling test set to obtain the Mura defect detection model.

[0046] The present invention can improve the quality and clarity of the image by performing image enhancement on the image of the bright screen defect annotation of the historical display screen. Common image enhancement techniques include adjusting brightness, contrast, removing noise, enhancing edges, etc. Through image enhancement, the defect annotation image can be made clearer, the defect area can be highlighted, and subsequent defect detection and analysis can be facilitated. The label format conversion of the enhanced image of the bright screen defect annotation of the historical display screen is to mark the defect area in the image as a corresponding category label. The defect information in the image is matched with the corresponding label to generate a corresponding defect annotation label for training and evaluating the defect detection model. The enhanced historical display screen bright screen defect annotation label is divided into a modeling training set and a modeling test set according to a preset division ratio for model training and evaluation. The purpose of dividing the training set and the test set is to verify the generalization ability of the model and avoid overfitting the model on the training data. The modeling training set and the modeling test set are modeled using the YOLO3 algorithm, which can realize the detection and positioning of the bright screen defects of the display screen. YOLO3 is a target detection algorithm based on deep learning, which can realize real-time target detection tasks with high accuracy and precision. The Gaussian error evaluation iteration of the defect detection model is performed based on the modeling test set, aiming to optimize and improve the model. By comparing the difference between the model detection results and the labels, the model error can be calculated, and the model can be continuously optimized through an iterative process to improve the accuracy and robustness of the model.

[0047] Optionally, step S3 specifically includes:

[0048] Obtaining a display image set of the display screen to be detected;

[0049] Extracting the bright screen display image of the display screen to be detected from the display image set of the display screen to be detected, so as to obtain the bright screen display image of the display screen to be detected;

[0050] Extract features of the bright screen display image of the display screen to be detected, so as to obtain the bright screen display features of the display screen to be detected;

[0051] Obtain the display screen detection backlight source information and image shooting point location information;

[0052] The backlight information of the display screen is used to calculate the correlation with the bright screen display characteristics of the display screen to be detected, so as to obtain the display-backlight correlation coefficient;

[0053] The correlation calculation is performed using the image shooting point position information and the bright screen display characteristics of the display screen to be detected, so as to obtain the display-shooting point correlation coefficient;

[0054] Objective impact profiling is performed based on the display-backlight correlation coefficient and the display-shooting point correlation coefficient;

[0055] According to the objective impact portrait, display defect detection is performed on the bright screen display image of the display to be inspected through the Mura defect detection model, so as to obtain the first display performance detection data.

[0056] The present invention can obtain the original image data to be detected by acquiring the display image of the display screen to be detected. In the display image set of the display screen to be detected, a partial image of the display screen bright screen display is extracted, and the background and other irrelevant information are removed, so as to obtain the display image of the display screen to be detected. Feature extraction is performed on the bright screen display image of the display screen to be detected, and information representing the display features of the bright screen can be extracted. These features can be texture features, brightness features, edge features, etc., which are used to describe the bright screen display status of the display screen to be detected. Acquiring the backlight source information of the display screen detection can obtain information such as the state and brightness of the backlight source, which can be used as a reference for subsequent evaluation of display performance. At the same time, acquiring the image shooting point position information can record the position and angle when the image is shot, which is helpful for subsequent analysis and evaluation of defects of the display screen. By calculating the display-backlight source correlation coefficient, the relationship between the brightness of the display screen and the backlight source can be quantified. According to the size of the correlation coefficient, it can be evaluated whether the brightness of the display screen is affected by the backlight source and whether the backlight source is normal. By calculating the display-shooting point correlation coefficient, the relationship between the characteristics of the display screen image and the shooting point position can be quantified. This can help determine whether the angle and position of the image will affect the observation of the display screen. An objective impact portrait is generated based on the display-backlight correlation coefficient and the display-shooting point correlation coefficient. This portrait can reflect the impact of factors such as the backlight and shooting point position on the display observation, which helps to further analyze the performance of the display. Using the generated objective impact portrait as input, the Mura defect detection model is used to perform defect detection on the bright screen display image of the display to be inspected. The Mura defect detection model can identify and locate defects on the display, such as uneven brightness, mottled, dark spots, etc.

[0057] Optionally, step S4 specifically includes:

[0058] Extracting the off-screen display image from the display image set of the display screen to be detected, thereby obtaining the off-screen display image of the display screen;

[0059] The screen-off display image is segmented by a display screen image segmentation algorithm, thereby obtaining a screen-off display segmentation image;

[0060] Obtain infrared images of the display screen when it is turned off by infrared camera technology;

[0061] Performing surface foreign matter detection on the screen-off display segmentation image of the screen-off display according to the screen-off infrared image of the screen-off display, thereby obtaining surface foreign matter data;

[0062] Calculating the display screen area of ​​the display image to be detected, thereby obtaining display screen area data;

[0063] The anti-surface foreign matter interference performance is calculated based on the surface foreign matter data and the display screen area data, thereby obtaining the second display screen performance test data.

[0064] The present invention processes the display image set of the display screen to be detected to extract the off-screen display image of the display screen. This helps to obtain the display screen image in the off-screen state for subsequent analysis and processing. The off-screen display image of the display screen is segmented using a display screen image segmentation algorithm. The display screen image can be segmented into different areas for subsequent analysis and detection. The off-screen infrared image of the display screen is obtained by infrared camera technology. The infrared image can provide the thermal distribution of the display screen in the off-screen state, which helps to detect possible problems. The off-screen infrared image of the display screen is used to detect surface foreign matter on the off-screen display segmented image of the display screen. This can help detect whether there are foreign matter, stains or other abnormal conditions on the surface of the display screen. The display image of the display screen to be detected is calculated for area to obtain the actual display area of ​​the display screen. This information can be used for subsequent calculation and evaluation of the performance of the display screen. Combined with the surface foreign matter data and the display area data, the anti-surface foreign matter interference calculation is performed. This can quantitatively evaluate the performance of the display screen when there are surface foreign matter, and help determine whether the performance of the display screen is interfered.

[0065] Optionally, step S5 specifically includes:

[0066] The first display screen performance test data and the second display screen performance test data are calculated by using a display screen comprehensive evaluation coefficient calculation formula to obtain a display screen performance evaluation coefficient to be tested, wherein the display screen comprehensive evaluation coefficient calculation formula is specifically:

[0067]

[0068] Where EC is the comprehensive evaluation coefficient of the display screen, D is the bright screen defect coefficient, B is the reference brightness of the display screen, C is the contrast level, V is the observation viewing angle coefficient, A is the surface foreign matter interference factor, R is the surface foreign matter interference weight coefficient, n is the number of observations, M is the average deviation of the observations, N is the total number of minor influencing factors, i is the minor influencing factor index, P is the total number of minor influencing factors, and i is the parameter of the ith secondary influencing factor, Q i is the weight of the ith secondary influencing factor, T is the time point, f(t) is the time factor change function, and e is the base of the natural logarithm;

[0069] The present invention constructs a display screen comprehensive evaluation coefficient calculation formula for calculating the first display screen performance test data and the second display screen performance test data. The formula fully considers the bright screen defect coefficient D, the display screen reference brightness B, the contrast level C, the observation viewing angle coefficient V, the surface foreign matter interference factor A, the surface foreign matter interference weight coefficient R, the number of observations n, the observed average deviation M, the total number of secondary influencing factors N, the secondary influencing factor index i, and the parameter P of the ith secondary influencing factor. i , the weight of the i-th secondary influencing factor Q i , time point T, time factor change function f(t), natural logarithm base e, form a functional relationship:

[0070]

[0071] in Represents the logarithm of the square root of the ratio of the brightness defect data to the reference brightness level. The contrast ratio is the rate of change of contrast relative to viewing angle, which measures the rate of change of contrast when the viewing angle changes. It is an exponential function of the ratio of the surface foreign body interference factor A and the weight coefficient R to the contrast C. It is the nth root of the square root of the observed mean deviation M. is the secondary influencing factor parameter P i and weight Q i The sum of the products of represents the comprehensive impact of other minor factors on the performance of the display. is the integral of the function f(t) in the time range [0,T], taking into account the effect of time on the performance of the display. The part represents the compensation for the brightness defect. By calculating the logarithm of the square root of the ratio of the brightness defect to the reference brightness level, it can be quantified and adjusted accordingly to improve the brightness performance of the display. The term represents the rate of change of contrast when the viewing angle changes. The existence of this part can help to perceive the change of contrast at different viewing angles, so as to make a more comprehensive evaluation and adjustment of the display. The term represents the exponential function of the ratio of the surface foreign matter interference factor and the weight coefficient to the contrast ratio. This part can be used to correct the adverse effects of foreign matter interference on the display performance, which may include reducing noise and light scattering. and Represent the observed mean deviation and the combined effect of the parameters and weights of the secondary influencing factors, respectively. The existence of these parts can comprehensively consider multiple aspects and factors of the display performance and weigh their impact according to their importance. The term represents the integral of the function f(t) over time, which is used to consider the effect of time on the performance of the display. This means that the formula can be used to analyze the timing changes in the performance of the display and make adjustments or predictions based on the time factor.

[0072] Merging the first display screen performance test data and the second display screen performance test data to obtain the display screen performance test data to be tested;

[0073] The performance evaluation data of the display screen to be detected is parameter-marked according to the performance evaluation coefficient of the display screen to be detected, so as to obtain the performance evaluation data of the display screen to be detected.

[0074] The present invention uses a display screen comprehensive evaluation coefficient calculation formula, combined with the first display screen performance test data and the second display screen performance test data, to calculate the performance evaluation coefficient of the display screen to be tested. The evaluation coefficient comprehensively considers the weights of multiple performance indicators and can quantitatively reflect the overall performance level of the display screen to be tested. Data merging is performed on the first display screen performance test data and the second display screen performance test data. The two data sets can be merged into a complete data set for subsequent analysis and processing. Through data merging, the performance of the display screen to be tested can be more comprehensively understood and a comprehensive evaluation can be performed. According to the performance evaluation coefficient of the display screen to be tested, the performance test data of the display screen to be tested is parameter-labeled. This can help classify and identify parameters of different performance levels so as to better understand and analyze performance data. Parameter labeling can indicate which parameters perform well and which parameters need to be improved or adjusted.

[0075] Optionally, step S6 specifically includes:

[0076] Building a display screen main body model according to the display image set of the display screen to be detected;

[0077] Extracting features from the performance evaluation data of the display screen to be tested, thereby obtaining the performance features of the display screen to be tested;

[0078] Constructing a display screen shadow model according to the performance characteristics of the display screen to be tested;

[0079] The display screen main body model and the display screen shadow model are combined based on the convolutional neural network to obtain a display screen combination model;

[0080] The display screen combination model is optimized by minimizing the difference parameters through supervised learning method, so as to obtain the display screen digital twin model;

[0081] Upload the digital twin model of the display to the visualization platform to perform 3D visualization tasks of display defects.

[0082] According to the present invention, a display screen main body model can be constructed according to the image displayed by the display screen to be detected. This model can be used to represent the appearance and shape of the display screen, so as to better understand the physical structure of the display screen and its association with other components. By extracting features from the performance evaluation data of the display screen to be detected, useful key features that can describe the performance of the display screen can be extracted from the original data. These features can include indicators such as brightness, contrast, color accuracy, response time, and other information related to display quality. According to the performance characteristics of the display screen to be detected, a display screen shadow model can be constructed. This model is an abstraction and representation of the performance of the display screen, and can capture the key performance characteristics of the display screen for subsequent analysis and processing. Based on the convolutional neural network, the display screen main body model and the display screen shadow model are combined to construct a comprehensive display screen combination model. By combining these two models, the appearance information and performance characteristics can be comprehensively considered, so as to more comprehensively describe the performance of the display screen to be detected. Through the supervised learning method, on the basis of the display screen combination model, the model is optimized to minimize the difference parameters. This can improve the accuracy and adaptability of the model, so that the display screen digital twin model better reflects the performance of the real display screen. The optimized display screen digital twin model is uploaded to the visualization platform, and the 3D visualization task of the display screen defects can be performed. This allows for intuitive observation and analysis of display defects, helping to provide more accurate diagnosis and evaluation, as well as guidance for improving and optimizing display performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0084] Figure 1 It is a schematic diagram of the steps of the performance detection method of the display screen of the present invention;

[0085] Figure 2 Detailed step flow diagram of step S1 in the present invention;

[0086] Figure 3 Detailed step flow diagram of step S13 in the present invention;

[0087] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0088] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0089] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0090] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0091] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for detecting the performance of a display screen, the method comprising the following steps:

[0092] Step S1: obtaining a historical display screen bright screen display image, performing defect detection on the historical display screen bright screen display image, thereby obtaining a historical display screen bright screen defect annotated image;

[0093] In this embodiment, a series of historical display screen bright screen display images are obtained from a database or image library, and defect detection is performed on these images using image processing and computer vision algorithms. For example, edge detection, threshold segmentation, morphological operation and other technologies can be used to detect and mark display screen defects in images to generate historical display screen bright screen defect annotated images.

[0094] Step S2: constructing a model for the historical display screen bright screen defect annotated images based on a deep learning method, thereby obtaining a Mura defect detection model;

[0095] In this embodiment, deep learning technology, such as YOLO3, is used to build a mura defect detection model using historical display screen defect annotated images as training data. Through multiple iterations of training and optimization of model parameters, the model can accurately detect display screen defects.

[0096] Step S3: obtaining a display image set of the display screen to be inspected, and performing a first display screen performance test on the display image set of the display screen to be inspected using a Mura defect detection model, thereby obtaining first display screen performance test data;

[0097] In this embodiment, a display image set of the display screen to be detected is obtained from a database or an image library, a bright screen display image is obtained from the display screen to be detected, and these images are input into the constructed Mura defect detection model. The model will perform defect detection on these images and output the first display screen performance detection data, such as defect type, defect location, defect degree and other information.

[0098] Step S4: extracting the off-screen display image from the display image set to be detected, thereby obtaining the off-screen display image of the display screen, and performing a second display screen performance test on the off-screen display image of the display screen, thereby obtaining second display screen performance test data;

[0099] In this embodiment, the display images in the off-screen state are obtained from the display screen to be detected. The off-screen display images are processed and analyzed to obtain the performance detection data of the second display screen, such as uniformity, brightness uniformity, grayscale response and other indicators.

[0100] Step S5: calculating a comprehensive evaluation coefficient for the first display screen performance test data and the second display screen performance test data, thereby obtaining a performance evaluation coefficient for the display screen to be tested, and merging the first display screen performance test data and the second display screen performance test data according to the performance evaluation coefficient for the display screen to be tested, thereby obtaining performance evaluation data for the display screen to be tested;

[0101] In this embodiment, according to certain evaluation algorithms and rules, the first display screen performance test data and the second display screen performance test data are combined to calculate a comprehensive evaluation coefficient. The coefficient can reflect the overall performance level of the display screen. Then, according to the size of the evaluation coefficient, the first display screen performance test data and the second display screen performance test data are combined to generate performance evaluation data of the display screen to be tested.

[0102] Step S6: construct a digital twin model of the display screen according to the performance evaluation data of the display screen to be tested, and upload the digital twin model of the display screen to the visualization platform to perform the 3D visualization task of the display screen defects.

[0103] In this embodiment, the performance evaluation data of the display screen to be tested is used to construct a digital twin model of the display screen. The model may include information such as the geometric shape, material properties, and pixel distribution of the display screen. Then, the digital twin model is uploaded to the visualization platform, and the defects of the display screen are presented through 3D visualization technology, so that users can intuitively observe and analyze the problems of the display screen.

[0104] The present invention can obtain a historical display screen bright screen defect annotated image by acquiring a historical display screen bright screen display image and performing defect detection. Defect sample data is established, which contains various defects that may occur when the display screen is bright. These data will be used for the construction and training of the deep learning model in the subsequent process. The historical display screen bright screen defect annotated image is modeled based on the deep learning method to obtain a Mura defect detection model. A large number of bright screen defect annotated images are used, and the model is constructed by a deep learning algorithm to improve the accuracy and consistency of the detection and reduce the subjectivity. Through the deep learning model, various defects in the bright screen display image can be accurately detected, such as missing pixels, uneven brightness or abnormal colors. The Mura defect detection model can be used to detect defects in the display screen to be detected in the subsequent process. By using the Mura defect detection model to perform a first display screen performance test on the display image set of the display screen to be detected, the first display screen performance test data can be obtained. It is possible to promptly discover defects that may exist when the display screen to be detected is bright, and obtain quantitative data on the performance of the display screen. These data can help evaluate and judge the quality of the display screen. The display screen to be detected is subjected to off-screen display image extraction, and the off-screen display image is subjected to a second display screen performance test, thereby obtaining the second display screen performance test data. This can detect possible problems of the display screen to be tested in the off state, such as bad pixels, dark spots or uneven brightness. The performance test data of the second display screen can further improve the evaluation of the performance of the display screen. The display screen is tested in an automated manner, which greatly reduces the manpower and time costs and improves efficiency. The performance evaluation coefficient of the display screen to be tested can be obtained by calculating the comprehensive evaluation coefficient of the performance test data of the first display screen and the performance test data of the second display screen. Based on objective data calculation, the influence of subjective evaluation is reduced, and a more accurate performance evaluation result is provided. The data obtained from the two tests are comprehensively analyzed and weighed to obtain the evaluation result of the overall performance of the display screen. By calculating the evaluation coefficient, the performance status of the display screen to be tested can be more accurately evaluated. A digital twin model of the display screen is constructed according to the performance evaluation data of the display screen to be tested, and it is uploaded to the visualization platform to perform the 3D visualization task of the display screen defects. The performance data and defect information of the display screen are combined with the digital twin model to realize the visualization of the display screen defects. In this way, the defect problems of the display screen can be intuitively observed and analyzed, and corresponding measures can be taken to repair or improve them. Through the visual presentation of the digital model, the defect problems of the display screen are more intuitive and clear, reducing the deviation of individual subjective evaluation.

[0105] Optionally, step S1 specifically includes:

[0106] Step S11: Obtain the historical display screen bright screen display image;

[0107] In this embodiment, a series of historical display screen bright display images are obtained from historical records, a database or an image library as input data.

[0108] Step S12: converting the historical display screen bright screen display image into a grayscale image, thereby obtaining a display screen bright screen display grayscale image;

[0109] In this embodiment, an image processing library is used to convert the acquired historical display screen bright screen display image into a grayscale image, that is, the RGB value of each pixel is converted into a corresponding grayscale value.

[0110] Step S13: performing background segmentation on the grayscale image displayed on the bright screen of the display screen by using a display screen image segmentation algorithm, thereby obtaining a display foreground image;

[0111] In this embodiment, an image segmentation algorithm, such as threshold-based segmentation, edge detection, etc., is used to distinguish the foreground and background of the display screen. According to the threshold or rule selected by the segmentation algorithm, pixels in the image are assigned as foreground or background, thereby obtaining a display foreground image.

[0112] Step S14: binarizing the display foreground image to obtain a display binary image;

[0113] In this embodiment, a binarization algorithm (such as fixed threshold binarization or adaptive threshold binarization) is used to binarize the display foreground image. An appropriate threshold is set, the foreground pixels are set to white (255) and the background pixels are set to black (0), and a display binary image is generated.

[0114] Step S15: performing connected domain defect determination on the displayed binary image, thereby obtaining a historical display screen bright screen defect annotated image.

[0115] In this embodiment, a connected domain analysis algorithm, such as connected region marking, erosion and dilation techniques, is used to identify connected regions in the binary display image. The connected regions in the binary display image are detected and analyzed to identify possible display screen defects, and the identification results are marked in the historical display screen bright screen image to generate a historical display screen bright screen defect marked image.

[0116] The present invention obtains the bright screen display images of historical display screens, and collects and obtains a large number of bright screen display images of historical display screens. These images are the basic data required for subsequent detection and evaluation. By collecting a large number of images, a rich data set can be established, which is helpful for model training and performance improvement. The bright screen display images of the historical display screens are converted into grayscale images, which simplifies the complexity of subsequent processing while retaining the key information of the image. The grayscale image only contains brightness information, without considering the color. This can reduce the resource consumption of calculation and storage, and is more convenient for image processing and analysis. The bright screen display grayscale image of the display screen is segmented by the display screen image segmentation algorithm, and the display foreground image can be obtained by segmentation, and the interference of the background is removed, so that the subsequent processing is more focused on the display content, and the accuracy and reliability of defect detection are improved. The binarization process converts the image into a binary image containing only black and white colors. The image can be further simplified, making the subsequent defect detection simpler and more efficient. The white in the binary image represents the display content, and the black represents the non-display content, so that the edges and details of the defects are clearer and more prominent. Conduct connected domain defect determination on binary images; connected domain defect determination refers to determining the defective parts by analyzing and determining the connected areas in binary images. Locate and annotate defective areas in historical display screens to provide annotation data for subsequent model training and detection. These annotated images can be used to train deep learning models to improve the accuracy and robustness of defect detection.

[0117] Optionally, step S13 is specifically:

[0118] Step S131: extracting grayscale pixel values ​​of the grayscale image displayed on the bright screen of the display screen, thereby obtaining the grayscale pixel values ​​of the bright screen of the display screen;

[0119] In this embodiment, each pixel of the grayscale image displayed on the bright screen of the display screen is traversed, and the grayscale pixel value is extracted, thereby obtaining the grayscale pixel value of the bright screen of the display screen.

[0120] Step S132: performing statistical analysis on the grayscale pixel values ​​of the display screen to obtain a grayscale pixel value threshold;

[0121] In this embodiment, the extracted grayscale pixel values ​​are statistically analyzed, such as calculating the average value, standard deviation, histogram, etc. According to the statistical analysis result, an appropriate threshold is determined as the threshold of the grayscale pixel value.

[0122] Step S133: performing edge segmentation on the grayscale image displayed on the display screen by using a display screen image segmentation algorithm according to the grayscale pixel value threshold, thereby obtaining a display screen edge segmentation grayscale image;

[0123] In this embodiment, an edge detection algorithm, such as the Canny edge detection algorithm, is used to perform edge detection on the bright screen grayscale image of the display screen. The edge detection result image is binarized according to the grayscale pixel value threshold to obtain a display screen edge segmentation grayscale image.

[0124] Step S134: Gaussian blurring the segmented grayscale image at the edge of the display screen to obtain an image to be background-eliminated;

[0125] In this embodiment, a Gaussian filter is used to perform blur processing on the grayscale image of the edge segmentation of the display screen. Gaussian blur can reduce noise and smooth the image to better eliminate the background.

[0126] Step S135: performing background removal on the image to be background-removed using a background removal method based on the domain information, thereby obtaining a display foreground image.

[0127] In this embodiment, a background elimination algorithm based on domain information is used, such as threshold segmentation based on neighboring pixels, segmentation based on mean shift, etc. By analyzing the neighboring relationship and color value of pixels, the background and foreground in the image are distinguished to obtain a displayed foreground image.

[0128] The present invention extracts grayscale pixel values ​​from a grayscale image displayed on a bright screen of a display screen; converts a color image into a grayscale image and extracts grayscale pixel values. Grayscale pixel values ​​reflect the brightness information of each pixel in an image. By extracting grayscale pixel values, subsequent statistical analysis and processing can be performed more conveniently. Statistical analysis is performed on the grayscale pixel values ​​of the bright screen of the display screen; the beneficial effect of statistically analyzing the grayscale pixel values ​​is to determine a suitable grayscale pixel value threshold value according to the brightness distribution characteristics of the image. The selection of the threshold value is based on the brightness difference between the foreground and the background in the image, which will help the subsequent image segmentation to accurately distinguish the display foreground from the background. The edge segmentation of the grayscale image displayed on the bright screen of the display screen is performed by a display screen image segmentation algorithm according to the grayscale pixel value threshold value; image segmentation is performed based on the grayscale pixel value threshold value, and the image is divided into the foreground and the background. Through edge segmentation, the edge information of the display screen can be obtained, which is very helpful for further processing and analysis of the display foreground and background areas. Gaussian blur is performed on the grayscale image segmented by the edge of the display screen; Gaussian blur is a commonly used image processing operation that can reduce noise and interference in the image and smooth the detailed information of the image. By performing Gaussian blur on the edge segmentation grayscale image, the edge of the display foreground can be made smoother and more accurate image data can be provided for subsequent background removal. The background removal method based on domain information removes the background of the background-removed image; the background removal method based on domain information removes the background part of the image by processing the background-removed image. Background removal extracts the foreground part of the display screen, so that subsequent defect detection and analysis can focus on the display content, improving the accuracy and reliability of detection.

[0129] Optionally, the function formula of the display screen image segmentation algorithm in step S133 is specifically:

[0130]

[0131] Where E is the edge intensity value, I is the gray value of the pixel, S is the gray pixel value threshold, k is the edge enhancement coefficient, α is the image enhancement index, β is the translation parameter, and γ is the exponential parameter. is the gradient operator, x is the horizontal coordinate of the pixel in the image, y is the vertical coordinate of the pixel in the image, and e is the base of the natural logarithm.

[0132] The present invention constructs a function formula of a display screen image segmentation algorithm, which is used to calculate the edge intensity value of a grayscale image displayed on a bright screen of a display screen. The formula fully considers the grayscale value I of the pixel that affects the edge intensity value E, the grayscale pixel value threshold S, the edge enhancement coefficient k, the image enhancement index α, the translation parameter β, the exponential parameter γ, the gradient operator The horizontal coordinate x of the pixel in the image, the vertical coordinate y of the pixel in the image, and the base e of the natural logarithm form a functional relationship:

[0133]

[0134] in Represents the gradient operator, which is used to calculate the gradient of the image. Application Compute gradient information. represents the Euclidean norm, which calculates the magnitude of the gradient. Where x and y represent the gradient values ​​in the corresponding directions, and β is a translation parameter. By adding β inside the square root, the curve is translated to adapt to the gradient value distribution of different images. γ represents the exponential parameter, which is used to adjust By adjusting γ, the response of edge enhancement to the gradient value can be controlled. I is the gray value of the pixel, which represents the input gray image. S represents the gray pixel value threshold, which is used to divide the gray image into two parts: bright and dark. The edge will be generated in the area where the brightness is higher than the threshold. k represents the edge enhancement coefficient, which is used to adjust the strength of the edge. Increasing k will make the edge more obvious. α represents the image enhancement index, which is used to adjust the effect of brightness on the edge. Increasing α will enhance the contribution of brightness to the edge. Used to adjust the impact of edge enhancement. The magnitude of this function decreases as the brightness of the input grayscale image increases. is a logic function used to limit the brightness range of edge generation. Larger γ and (SI) will reduce the brightness of edge generation. represents the calculation of gradient, and generates edge image through gradient operation. Then, the edge image is Multiply to adjust the edge strength and brightness. Finally, multiply by Used to control the brightness range of edge generation. The gradient operation and edge strength adjustment parts in the formula can effectively extract the edge information of the image. By calculating the gradient and adjusting the edge strength, the edge of the image can be separated from the background, thereby realizing applications such as image segmentation and target detection. The edge enhancement part in the formula can enhance the edges in the image, making them more obvious and clear. By adjusting the edge enhancement coefficient and the image enhancement index, the edge strength and the degree of response to brightness can be controlled, thereby improving the visual effect and recognition of the image. The formula introduces multiple parameters, such as the translation parameter β, the exponential parameter γ, and the grayscale pixel value threshold S. These parameters can be adjusted according to the specific application and image characteristics to adapt to different image scenes and needs. By adjusting these parameters, the performance and applicability of the algorithm can be optimized to achieve better image segmentation effects. By introducing logical function terms in the formula The brightness range of edge generation can be limited to make the edges of darker areas more obvious. This can increase the contrast of the image and improve visual perception.

[0135] Optionally, step S15 is specifically:

[0136] Denoising the binary image based on morphological filtering to obtain a denoised binary image.

[0137] In this embodiment, a morphological filtering method, such as an opening operation or a closing operation, is used to perform denoising on the binary image. The opening operation can eliminate small-sized noise points, and the closing operation can fill small-sized holes. These morphological operations will make the denoised binary image smoother and remove unnecessary details and noise.

[0138] Calculate the connected domain area of ​​the denoised binary image to obtain the connected domain area data;

[0139] In this embodiment, the connected domain analysis method is used to process the denoised binary image to calculate the area of ​​each connected domain. A connected domain refers to an area composed of adjacent pixels. By calculating the number of pixels in each connected domain, the area data of the connected domain can be obtained.

[0140] The spectrum data of the connected domain area is converted into spectrum data by using a one-dimensional discrete Fourier transform method, so as to obtain the spectrum data of the connected domain area;

[0141] In this embodiment, one-dimensional discrete Fourier transform (DFT) is applied to the connected domain area data for spectrum conversion. DFT converts the area data from the time domain to the frequency domain to obtain the connected domain area spectrum data.

[0142] Perform abnormal band detection on the connected domain area spectrum data to obtain the abnormal band of the connected domain area;

[0143] In this embodiment, abnormality detection is performed on the connected domain area spectrum data to find abnormal bands or frequency components. Abnormal band detection can be performed using statistical methods, filtering methods, or threshold methods. Abnormal bands indicate abnormal area values ​​in the spectrum, which may indicate defects in the display screen.

[0144] The image of the abnormal band of the connected domain area is converted by the inverse Fourier transform method to obtain the bright screen defect image;

[0145] In this embodiment, the inverse Fourier transform (IDFT) is applied to the abnormal band of the connected domain area to convert the spectrum data back to the image domain. IDFT restores the frequency domain data to a bright screen defect image, in which the brightness value represents the degree of abnormality.

[0146] Defects of the denoised binary image are annotated according to the bright screen defect image, so as to obtain a bright screen defect annotated image of a historical display screen.

[0147] In this embodiment, the bright screen defect image is used as a reference to mark the defect area on the binary image after denoising. The defect area can be marked by using a boundary box, color marking, etc. The marked image is the historical display screen bright screen defect marked image.

[0148] The present invention performs denoising on a binary image based on a denoising method of morphological filtering; morphological filtering is an effective image denoising method, which can remove noise in an image and retain the main target structure by dilating and corroding structural elements in an image. The application of the morphological filtering method to denoise a binary image can eliminate noise points and small discontinuous areas in the image, thereby obtaining a clear and continuous denoised binary image. The connected domain area is calculated for the denoised binary image; after the denoised binary image is obtained, the size information of each connected domain can be obtained by calculating the area of ​​the connected domain, that is, the size of the adjacent pixel set. This helps to count the size distribution of different connected domains in the image and obtain meaningful image attributes therefrom. The connected domain area data is converted into spectrum data by a one-dimensional discrete Fourier transform method; the connected domain area data can be converted into spectrum data by performing DFT processing on the connected domain area data. Such spectrum data can better reveal the spectrum characteristics of the connected domain area distribution, such as frequency components, intensity and other information. The connected domain area spectrum data is detected by abnormal bands; the connected domain area spectrum data is detected by abnormal bands, and abnormal bands in the spectrum can be identified. These abnormal bands may correspond to special structures or noise in the image. By detecting and analyzing these abnormal bands, connected domains with special properties or abnormal characteristics can be extracted. The abnormal bands of the connected domain area are converted into images by the inverse Fourier transform method; the spectrum data extracted from the abnormal bands can be converted back to the spatial domain image by applying the inverse Fourier transform. In this way, a bright screen defect image can be obtained, which contains areas with abnormal brightness, which is helpful for further analysis and processing. Defects are annotated on the denoised binary image based on the bright screen defect image; based on the bright screen defect image, the corresponding defect area can be annotated to visualize and locate it in the image. This helps to quickly identify and analyze bright screen defects on historical display screens, and perform subsequent processing and repair.

[0149] Optionally, step S2 specifically includes:

[0150] Performing image enhancement on the historical display screen bright screen defect annotated image, thereby obtaining an enhanced historical display screen bright screen defect annotated image;

[0151] In this embodiment, image processing techniques are used to enhance the image, such as adjusting brightness, contrast, sharpening, etc. Some filters or enhancement algorithms may also be applied, such as histogram equalization, adaptive histogram equalization, enhancement filter, etc. Image enhancement aims to improve the visual quality of the image and make defects easier to detect and identify.

[0152] The enhanced historical display screen bright screen defect annotation image is converted into a label format, thereby obtaining an enhanced historical display screen bright screen defect annotation label;

[0153] The label format conversion in this embodiment generally involves labeling the defect area in the image as a specific label or category. An image labeling tool or script can be used to manually or automatically label the defect location as a rectangular box, mask, or pixel-level annotation. The purpose of the label format conversion is to associate the defect location information in the image with the corresponding label data.

[0154] The enhanced historical display screen bright screen defect annotation labels are divided into a modeling training set and a modeling test set according to a preset division ratio;

[0155] In this embodiment, the enhanced image and the corresponding label are divided according to a certain ratio according to the ratio of 7:3 or 8:2. Common division methods include random division, division by category, etc. The purpose of dividing the modeling training set and the modeling test set is to use the training set to learn and adjust the model parameters during training, and then evaluate the performance of the model on the test set.

[0156] Model the modeling training set and the modeling test set based on the YOLO3 algorithm to obtain a defect detection model;

[0157] In this embodiment, the YOLO3 algorithm is used to train the target detection model. The modeling training set is used as input data, and the model parameters are optimized through the training algorithm. The model will learn how to detect and locate defect areas in the image.

[0158] The Gaussian error evaluation iteration is performed on the defect detection model according to the modeling test set to obtain the Mura defect detection model.

[0159] In this embodiment, the trained defect detection model is evaluated using the modeling test set. The Gaussian error between the predicted result and the true label, such as the IoU (intersection over union) and other indicators, can be calculated. Based on the evaluation results, the model can be adjusted and iterated to improve the performance and accuracy of the defect detection model.

[0160] The present invention can improve the quality and clarity of the image by performing image enhancement on the image of the bright screen defect annotation of the historical display screen. Common image enhancement techniques include adjusting brightness, contrast, removing noise, enhancing edges, etc. Through image enhancement, the defect annotation image can be made clearer, the defect area can be highlighted, and subsequent defect detection and analysis can be facilitated. The label format conversion of the enhanced image of the bright screen defect annotation of the historical display screen is to mark the defect area in the image as a corresponding category label. The defect information in the image is matched with the corresponding label to generate a corresponding defect annotation label for training and evaluating the defect detection model. The enhanced historical display screen bright screen defect annotation label is divided into a modeling training set and a modeling test set according to a preset division ratio for model training and evaluation. The purpose of dividing the training set and the test set is to verify the generalization ability of the model and avoid overfitting the model on the training data. The modeling training set and the modeling test set are modeled using the YOLO3 algorithm, which can realize the detection and positioning of the bright screen defects of the display screen. YOLO3 is a target detection algorithm based on deep learning, which can realize real-time target detection tasks with high accuracy and precision. The Gaussian error evaluation iteration of the defect detection model is performed based on the modeling test set, aiming to optimize and improve the model. By comparing the difference between the model detection results and the labels, the model error can be calculated, and the model can be continuously optimized through an iterative process to improve the accuracy and robustness of the model.

[0161] Optionally, step S3 specifically includes:

[0162] Obtaining a display image set of the display screen to be detected;

[0163] In this embodiment, an image acquisition device (such as a video camera or a camera) is used to acquire an image of the display screen to be inspected, and the clarity and accuracy of the display screen image shooting must be ensured.

[0164] Extracting the bright screen display image of the display screen to be detected from the display image set of the display screen to be detected, so as to obtain the bright screen display image of the display screen to be detected;

[0165] In this embodiment, image processing technology is used to extract the bright screen display image of the display screen in the display image set to be detected, remove background noise or other interference, and ensure that the extracted image area is consistent with the bright screen display of the display screen.

[0166] Extract features of the bright screen display image of the display screen to be detected, so as to obtain the bright screen display features of the display screen to be detected;

[0167] In this embodiment, computer vision technology, such as feature extraction algorithms (such as SIFT, SURF, HOG, etc.), is used to extract key features from the bright screen display image. The features can be local descriptors, texture features, color histograms, etc. of the image.

[0168] Obtain the display screen detection backlight source information and image shooting point location information;

[0169] In this embodiment, the display screen detection device or sensor is used to obtain the backlight information of the display screen, such as backlight intensity, color temperature, etc. The position information of the image shooting point can be obtained by recording the position of the image acquisition device or using a position sensor.

[0170] The backlight information of the display screen is used to calculate the correlation with the bright screen display characteristics of the display screen to be detected, so as to obtain the display-backlight correlation coefficient;

[0171] In this embodiment, a correlation analysis method, such as the Pearson correlation coefficient, is used to calculate the correlation between the bright screen display feature of the display screen and the backlight source information. The value of the correlation coefficient can indicate the degree of dependence of the bright screen feature of the display screen on the backlight source.

[0172] The correlation calculation is performed using the image shooting point position information and the bright screen display characteristics of the display screen to be detected, so as to obtain the display-shooting point correlation coefficient;

[0173] In this embodiment, a correlation analysis method is used to calculate the correlation between the bright screen display characteristics of the display screen and the position of the image shooting point. The value of the correlation coefficient can indicate the influence of shooting points at different positions on the bright screen display characteristics of the display screen.

[0174] Objective impact profiling is performed based on the display-backlight correlation coefficient and the display-shooting point correlation coefficient;

[0175] In this embodiment, the display-backlight correlation coefficient and the display-shooting point correlation coefficient are comprehensively considered to analyze the degree to which the bright screen characteristics of the display screen are affected by the backlight source and the shooting point position. An objective impact image can be generated to represent the impact of the backlight source and the shooting point on the bright screen display characteristics of the display screen.

[0176] According to the objective impact portrait, display defect detection is performed on the bright screen display image of the display to be inspected through the Mura defect detection model, so as to obtain the first display performance detection data.

[0177] In this embodiment, the objective influence portrait obtained previously is used, which reflects the influence of the backlight source and the shooting point position on the bright screen display characteristics of the display screen. The bright screen display image of the display screen to be tested is superimposed on the objective influence portrait at the pixel level to take the objective influence into account in the image. A bright screen display image of the display screen to be tested taking into account the objective influence is obtained, which reflects the influence of the actual backlight source and the shooting point position on the display effect. The Mura defect detection model is used to perform defect detection on the bright screen display image of the display screen to be tested taking into account the objective influence. The defect detection model analyzes the image to detect possible Mura defects, such as dark spots, bright spots, white spots, black spots, etc. Through the output of the model, the first display performance detection data is obtained, including information such as the type, location and severity of the detected defects.

[0178] The present invention can obtain the original image data to be detected by acquiring the display image of the display screen to be detected. In the display image set of the display screen to be detected, a partial image of the display screen bright screen display is extracted, and the background and other irrelevant information are removed, so as to obtain the display image of the display screen to be detected. Feature extraction is performed on the bright screen display image of the display screen to be detected, and information representing the display features of the bright screen can be extracted. These features can be texture features, brightness features, edge features, etc., which are used to describe the bright screen display status of the display screen to be detected. Acquiring the backlight source information of the display screen detection can obtain information such as the state and brightness of the backlight source, which can be used as a reference for subsequent evaluation of display performance. At the same time, acquiring the image shooting point position information can record the position and angle when the image is shot, which is helpful for subsequent analysis and evaluation of defects of the display screen. By calculating the display-backlight source correlation coefficient, the relationship between the brightness of the display screen and the backlight source can be quantified. According to the size of the correlation coefficient, it can be evaluated whether the brightness of the display screen is affected by the backlight source and whether the backlight source is normal. By calculating the display-shooting point correlation coefficient, the relationship between the characteristics of the display screen image and the shooting point position can be quantified. This can help determine whether the angle and position of the image will affect the observation of the display screen. An objective impact portrait is generated based on the display-backlight correlation coefficient and the display-shooting point correlation coefficient. This portrait can reflect the impact of factors such as the backlight and shooting point position on the display observation, which helps to further analyze the performance of the display. Using the generated objective impact portrait as input, the Mura defect detection model is used to perform defect detection on the bright screen display image of the display to be inspected. The Mura defect detection model can identify and locate defects on the display, such as uneven brightness, mottled, dark spots, etc.

[0179] Optionally, step S4 is specifically:

[0180] Extracting the off-screen display image from the display image set of the display screen to be detected, thereby obtaining the off-screen display image of the display screen;

[0181] In this embodiment, the screen-off display image in the display image set to be detected is extracted to generate the screen-off display image.

[0182] The screen-off display image is segmented by a display screen image segmentation algorithm, thereby obtaining a screen-off display segmentation image;

[0183] In this embodiment, a display screen image segmentation algorithm is used to segment the off-screen display image into different areas, each area representing a different display element or area.

[0184] Obtain infrared images of the display screen when it is turned off by infrared camera technology;

[0185] In this embodiment, an infrared camera or a camera device with an infrared sensor is used to capture an infrared image of the display screen when the display screen is off. Infrared camera technology can obtain the thermal distribution of the display screen when the display screen is off, which presents different information from the visible light image.

[0186] Performing surface foreign matter detection on the screen-off display segmentation image of the screen-off display according to the screen-off infrared image of the screen-off display, thereby obtaining surface foreign matter data;

[0187] In this embodiment, the infrared image is aligned with the segmented image of the display screen when the screen is off to ensure the matching of the corresponding areas. A surface foreign body detection algorithm, such as an algorithm based on infrared reflection characteristics or an algorithm based on deep learning, is used to detect and identify the presence of surface foreign bodies on the segmented image. The location, shape, size and other information of each detected foreign body are recorded to obtain the surface foreign body data.

[0188] Calculating the display screen area of ​​the display image to be detected, thereby obtaining display screen area data;

[0189] In this embodiment, image processing techniques, such as edge detection or shape-based analysis, are used to process the explicit image of the display screen to be detected. Based on the processed image, the effective area or outline of the display screen is calculated and its area is determined. The area data of the display screen is obtained for subsequent calculation and analysis.

[0190] The anti-surface foreign matter interference performance is calculated based on the surface foreign matter data and the display screen area data, thereby obtaining the second display screen performance test data.

[0191] In this embodiment, the surface foreign matter data and the display screen area data are combined to calculate the display screen's anti-surface foreign matter interference performance. For example, the degree of influence of the display screen by the surface foreign matter can be evaluated based on the relative relationship between the position, shape and size of the foreign matter and the display screen area. The second display screen performance detection data is output, which reflects the display screen's ability to resist surface foreign matter.

[0192] The present invention processes the display image set of the display screen to be detected to extract the off-screen display image of the display screen. This helps to obtain the display screen image in the off-screen state for subsequent analysis and processing. The off-screen display image of the display screen is segmented using a display screen image segmentation algorithm. The display screen image can be segmented into different areas for subsequent analysis and detection. The off-screen infrared image of the display screen is obtained by infrared camera technology. The infrared image can provide the thermal distribution of the display screen in the off-screen state, which helps to detect possible problems. The off-screen infrared image of the display screen is used to detect surface foreign matter on the off-screen display segmented image of the display screen. This can help detect whether there are foreign matter, stains or other abnormal conditions on the surface of the display screen. The display image of the display screen to be detected is calculated for area to obtain the actual display area of ​​the display screen. This information can be used for subsequent calculation and evaluation of the performance of the display screen. Combined with the surface foreign matter data and the display area data, the anti-surface foreign matter interference calculation is performed. This can quantitatively evaluate the performance of the display screen when there are surface foreign matter, and help determine whether the performance of the display screen is interfered.

[0193] Optionally, step S5 specifically includes:

[0194] The first display screen performance test data and the second display screen performance test data are calculated by using a display screen comprehensive evaluation coefficient calculation formula to obtain a display screen performance evaluation coefficient to be tested, wherein the display screen comprehensive evaluation coefficient calculation formula is specifically:

[0195]

[0196] Where EC is the comprehensive evaluation coefficient of the display screen, D is the bright screen defect coefficient, B is the reference brightness of the display screen, C is the contrast level, V is the observation viewing angle coefficient, A is the surface foreign matter interference factor, R is the surface foreign matter interference weight coefficient, n is the number of observations, M is the average deviation of the observations, N is the total number of minor influencing factors, i is the minor influencing factor index, P is the total number of minor influencing factors, and i is the parameter of the ith secondary influencing factor, Q i is the weight of the ith secondary influencing factor, T is the time point, f(t) is the time factor change function, and e is the base of the natural logarithm;

[0197] The present invention constructs a display screen comprehensive evaluation coefficient calculation formula for calculating the first display screen performance test data and the second display screen performance test data. The formula fully considers the bright screen defect coefficient D, the display screen reference brightness B, the contrast level C, the observation viewing angle coefficient V, the surface foreign matter interference factor A, the surface foreign matter interference weight coefficient R, the number of observations n, the observed average deviation M, the total number of secondary influencing factors N, the secondary influencing factor index i, and the parameter P of the ith secondary influencing factor. i, the weight of the i-th secondary influencing factor Q i , time point T, time factor change function f(t), natural logarithm base e, form a functional relationship:

[0198]

[0199] in Represents the logarithm of the square root of the ratio of the brightness defect data to the reference brightness level. The contrast ratio is the rate of change of contrast relative to viewing angle, which measures the rate of change of contrast when the viewing angle changes. It is an exponential function of the ratio of the surface foreign body interference factor A and the weight coefficient R to the contrast C. It is the nth root of the square root of the observed mean deviation M. is the secondary influencing factor parameter P i and weight Q i The sum of the products of represents the comprehensive impact of other minor factors on the performance of the display. is the integral of the function f(t) in the time range [0, T], taking into account the effect of time on the performance of the display. The part represents the compensation for the brightness defect. By calculating the logarithm of the square root of the ratio of the brightness defect to the reference brightness level, it can be quantified and adjusted accordingly to improve the brightness performance of the display. The term represents the rate of change of contrast when the viewing angle changes. The existence of this part can help to perceive the change of contrast at different viewing angles, so as to make a more comprehensive evaluation and adjustment of the display. The term represents the exponential function of the ratio of the surface foreign matter interference factor and the weight coefficient to the contrast ratio. This part can be used to correct the adverse effects of foreign matter interference on the display performance, which may include reducing noise and light scattering. and Represent the observed mean deviation and the combined effect of the parameters and weights of the secondary influencing factors, respectively. The existence of these parts can comprehensively consider multiple aspects and factors of the display performance and weigh their impact according to their importance. The term represents the integral of the function f(t) over time, which is used to consider the effect of time on the performance of the display. This means that the formula can be used to analyze the timing changes in the performance of the display and make adjustments or predictions based on the time factor.

[0200] Merging the first display screen performance test data and the second display screen performance test data to obtain the display screen performance test data to be tested;

[0201] In this embodiment, the first display performance test data and the second display performance test data are processed to ensure that the data format and unit are consistent. The two data sets are matched and merged based on a common identifier, such as a display serial number or a test timestamp. The merged data set will contain relevant parameters and indicators from the first and second display performance tests, providing comprehensive performance evaluation data for the display to be tested.

[0202] The performance evaluation data of the display screen to be detected is parameter-marked according to the performance evaluation coefficient of the display screen to be detected, so as to obtain the performance evaluation data of the display screen to be detected.

[0203] In this embodiment, each parameter in the performance test data of the display screen to be tested is multiplied by the corresponding evaluation coefficient to obtain the labeled evaluation data. The labeled evaluation data will reflect the performance of the display screen to be tested and weigh the importance of different parameters according to the evaluation coefficient.

[0204] The present invention uses a display screen comprehensive evaluation coefficient calculation formula, combined with the first display screen performance test data and the second display screen performance test data, to calculate the performance evaluation coefficient of the display screen to be tested. The evaluation coefficient comprehensively considers the weights of multiple performance indicators and can quantitatively reflect the overall performance level of the display screen to be tested. Data merging is performed on the first display screen performance test data and the second display screen performance test data. The two data sets can be merged into a complete data set for subsequent analysis and processing. Through data merging, the performance of the display screen to be tested can be more comprehensively understood and a comprehensive evaluation can be performed. According to the performance evaluation coefficient of the display screen to be tested, the performance test data of the display screen to be tested is parameter-labeled. This can help classify and identify parameters of different performance levels so as to better understand and analyze performance data. Parameter labeling can indicate which parameters perform well and which parameters need to be improved or adjusted.

[0205] Optionally, step S6 specifically includes:

[0206] Building a display screen main body model according to the display image set of the display screen to be detected;

[0207] In this embodiment, computer vision technology, such as feature extraction algorithms (such as SIFT, SURF, ORB, etc.) or deep learning models (such as convolutional neural networks) are used to extract features from the display image set. Based on the extracted features, a display screen main body model is established, which can describe the feature representation of the display screen to be detected.

[0208] Extracting features from the performance evaluation data of the display screen to be tested, thereby obtaining the performance features of the display screen to be tested;

[0209] In this embodiment, the performance evaluation data of the display screen to be detected is cleaned and normalized. According to the performance evaluation index of the display screen, a feature extraction method, such as statistical features, frequency domain features or other field-specific feature extraction methods, is used to extract features from the performance evaluation data of the display screen to be detected. The extracted features will represent the performance characteristics of the display screen to be detected.

[0210] Constructing a display screen shadow model according to the performance characteristics of the display screen to be tested;

[0211] In this embodiment, a machine learning or statistical modeling method is used to construct a display screen shadow model based on the performance characteristics of the display screen to be detected. The shadow model can be a linear model, a nonlinear model, or other models suitable for describing the performance characteristics of the display screen to be detected. The display screen shadow model will be used to predict the performance of the display screen to be detected under different conditions.

[0212] The display screen main body model and the display screen shadow model are combined based on the convolutional neural network to obtain a display screen combination model;

[0213] In this embodiment, the display screen main body model and shadow model are used as input, and a convolutional neural network (CNN) is used for model combination. CNN can learn the complex relationship and feature expression between different models. The CNN model is trained to integrate the information of the display screen main body model and shadow model to generate a display screen combination model.

[0214] The display screen combination model is optimized by minimizing the difference parameters through supervised learning method, so as to obtain the display screen digital twin model;

[0215] In this embodiment, supervised learning methods, such as back propagation algorithms, are used to train and optimize the display screen combination model. By minimizing the difference parameters, the display screen digital twin model can better predict the performance of the display screen to be tested. After training, a display screen digital twin model is obtained, which can provide accurate prediction and evaluation of the performance of the display screen to be tested.

[0216] Upload the digital twin model of the display to the visualization platform to perform 3D visualization tasks of display defects.

[0217] In this embodiment, the digital twin model of the display screen is integrated into the visualization platform, for example, 3D visualization is achieved using image processing and rendering technology. The relevant data of the display screen to be tested is input, and the performance is predicted and evaluated through the digital twin model. The visualization platform will generate 3D visualization graphics based on the prediction results of the digital twin model of the display screen to show the defects and performance characteristics of the display screen to be tested.

[0218] According to the present invention, a display screen main body model can be constructed according to the image displayed by the display screen to be detected. This model can be used to represent the appearance and shape of the display screen, so as to better understand the physical structure of the display screen and its association with other components. By extracting features from the performance evaluation data of the display screen to be detected, useful key features that can describe the performance of the display screen can be extracted from the original data. These features can include indicators such as brightness, contrast, color accuracy, response time, and other information related to display quality. According to the performance characteristics of the display screen to be detected, a display screen shadow model can be constructed. This model is an abstraction and representation of the performance of the display screen, and can capture the key performance characteristics of the display screen for subsequent analysis and processing. Based on the convolutional neural network, the display screen main body model and the display screen shadow model are combined to construct a comprehensive display screen combination model. By combining these two models, the appearance information and performance characteristics can be comprehensively considered, so as to more comprehensively describe the performance of the display screen to be detected. Through the supervised learning method, on the basis of the display screen combination model, the model is optimized to minimize the difference parameters. This can improve the accuracy and adaptability of the model, so that the display screen digital twin model better reflects the performance of the real display screen. The optimized display screen digital twin model is uploaded to the visualization platform, and the 3D visualization task of the display screen defects can be performed. This allows for intuitive observation and analysis of display defects, helping to provide more accurate diagnosis and evaluation, as well as guidance for improving and optimizing display performance.

[0219] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0220] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for detecting the performance of a display screen, characterized in that: The following steps are involved: Step S1: obtaining a historical display screen bright screen display image, performing defect detection on the historical display screen bright screen display image, thereby obtaining a historical display screen bright screen defect annotated image; wherein step S1 includes the following steps: Step S11: Obtain the historical display screen bright screen display image; Step S12: converting the historical display screen bright screen display image into a grayscale image, thereby obtaining a display screen bright screen display grayscale image; Step S13: performing background segmentation on the grayscale image displayed on the bright screen of the display screen by using a display screen image segmentation algorithm, thereby obtaining a display foreground image; wherein step S13 includes the following steps: Step S131: extracting grayscale pixel values ​​of the grayscale image displayed on the bright screen of the display screen, thereby obtaining the grayscale pixel values ​​of the bright screen of the display screen; Step S132: performing statistical analysis on the grayscale pixel values ​​of the display screen to obtain a grayscale pixel value threshold; Step S133: performing edge segmentation on the grayscale image displayed on the bright screen of the display screen by using a display screen image segmentation algorithm according to the grayscale pixel value threshold, thereby obtaining a display screen edge segmentation grayscale image; wherein the function formula of the display screen image segmentation algorithm is specifically: ； In the formula, E is the edge intensity value, I is the gray value of the pixel, S is the gray pixel value threshold, k is the edge enhancement coefficient, is the image enhancement index, is the translation parameter, is the exponential parameter, is the gradient operator, x is the horizontal coordinate of the pixel in the image, y is the vertical coordinate of the pixel in the image, and e is the base of the natural logarithm; Step S134: Gaussian blurring the segmented grayscale image at the edge of the display screen to obtain an image to be background-eliminated; Step S135: performing background removal on the background-removed image using a background removal method based on the domain information, thereby obtaining a display foreground image; Step S14: binarizing the display foreground image to obtain a display binary image; Step S15: performing connected domain defect determination on the displayed binary image, thereby obtaining a historical display screen bright screen defect annotated image; wherein step S15 includes the following steps: Denoising the binary image based on morphological filtering to obtain a denoised binary image. Calculate the connected domain area of ​​the denoised binary image to obtain the connected domain area data; The spectrum data of the connected domain area is converted into spectrum data by using a one-dimensional discrete Fourier transform method, so as to obtain the spectrum data of the connected domain area; Perform abnormal band detection on the connected domain area spectrum data to obtain the abnormal band of the connected domain area; The image of the abnormal band of the connected domain area is converted by the inverse Fourier transform method to obtain the bright screen defect image; Defect annotation is performed on the denoised binary image according to the bright screen defect image, thereby obtaining a bright screen defect annotation image of a historical display screen; Step S2: constructing a model for the historical display screen bright screen defect annotated images based on a deep learning method, thereby obtaining a Mura defect detection model; Step S3: obtaining a display image set of the display screen to be detected, and performing a first display screen performance test on the display image set of the display screen to be detected using a Mura defect detection model, thereby obtaining first display screen performance test data; wherein step S3 includes the following steps: Obtaining a display image set of the display screen to be detected; Extracting the bright screen display image of the display screen to be detected from the display image set of the display screen to be detected, so as to obtain the bright screen display image of the display screen to be detected; Extract features of the bright screen display image of the display screen to be detected, so as to obtain the bright screen display features of the display screen to be detected; Obtain the display screen detection backlight source information and image shooting point location information; The backlight source information of the display screen is used to calculate the correlation with the bright screen display characteristics of the display screen to be detected, so as to obtain the display-backlight source correlation coefficient; The correlation between the image shooting point position information and the bright screen display characteristics of the display screen to be detected is calculated to obtain the display-shooting point correlation coefficient; Objective impact profiling is performed based on the display-backlight correlation coefficient and the display-shooting point correlation coefficient; Performing display defect detection on the bright screen display image of the display screen to be inspected by using the Mura defect detection model according to the objective impact portrait, thereby obtaining first display screen performance detection data; Step S4: extracting the off-screen display image from the display image set to be detected, thereby obtaining the off-screen display image of the display screen, and performing a second display screen performance test on the off-screen display image of the display screen, thereby obtaining the second display screen performance test data; wherein step S4 includes the following steps: Extracting the off-screen display image from the display image set of the display screen to be detected, thereby obtaining the off-screen display image of the display screen; The screen-off display image is segmented by a display screen image segmentation algorithm, thereby obtaining a screen-off display segmentation image; Obtain infrared images of the display screen when it is turned off by infrared camera technology; Performing surface foreign matter detection on the screen-off display segmentation image of the screen-off display according to the screen-off infrared image of the screen-off display, thereby obtaining surface foreign matter data; Calculating the display screen area of ​​the display image to be detected, thereby obtaining display screen area data; Calculating the resistance to interference from surface foreign matter according to the surface foreign matter data and the display screen area data, thereby obtaining the second display screen performance test data; Step S5: calculating a comprehensive evaluation coefficient for the first display screen performance test data and the second display screen performance test data, thereby obtaining a performance evaluation coefficient for the display screen to be tested, and merging the first display screen performance test data and the second display screen performance test data according to the performance evaluation coefficient for the display screen to be tested, thereby obtaining performance evaluation data for the display screen to be tested; Step S6: constructing a display screen digital twin model according to the performance evaluation data of the display screen to be tested, and uploading the display screen digital twin model to the visualization platform to perform the display screen defect 3D visualization task; wherein step S6 includes the following steps: Building a display screen main body model according to the display image set of the display screen to be detected; Extracting features from the performance evaluation data of the display screen to be tested, thereby obtaining the performance features of the display screen to be tested; Constructing a display screen shadow model according to the performance characteristics of the display screen to be tested; The display screen main body model and the display screen shadow model are combined based on the convolutional neural network to obtain a display screen combination model; The display screen combination model is optimized by minimizing the difference parameters through supervised learning method, so as to obtain the display screen digital twin model; Upload the digital twin model of the display to the visualization platform to perform 3D visualization tasks of display defects.

2. The method for detecting the performance of a display screen according to claim 1, characterized in that: Step S2 is specifically as follows: Performing image enhancement on the historical display screen bright screen defect annotated image, thereby obtaining an enhanced historical display screen bright screen defect annotated image; The enhanced historical display screen bright screen defect annotation image is converted into a label format, thereby obtaining an enhanced historical display screen bright screen defect annotation label; The enhanced historical display screen bright screen defect annotation labels are divided into a modeling training set and a modeling test set according to a preset division ratio; Model the modeling training set and the modeling test set based on the YOLO3 algorithm to obtain a defect detection model; The Gaussian error evaluation iteration is performed on the defect detection model according to the modeling test set to obtain the Mura defect detection model.

3. The method for detecting the performance of a display screen according to claim 1, characterized in that: Step S5 is specifically as follows: The first display screen performance test data and the second display screen performance test data are calculated using a display screen comprehensive evaluation coefficient calculation formula to obtain a display screen performance evaluation coefficient to be tested, wherein the display screen comprehensive evaluation coefficient calculation formula is specifically: ； Where EC is the comprehensive evaluation coefficient of the display screen, D is the bright screen defect coefficient, B is the reference brightness of the display screen, C is the contrast level, V is the observation viewing angle coefficient, A is the surface foreign matter interference factor, R is the surface foreign matter interference weight coefficient, n is the number of observations, M is the average deviation of the observations, N is the total number of minor influencing factors, i is the minor influencing factor index, is the parameter of the ith secondary influencing factor, is the weight of the ith secondary influencing factor, T is the time point, is the time factor variation function, e is the base of natural logarithm; Merging the first display screen performance test data and the second display screen performance test data to obtain the display screen performance test data to be tested; The performance evaluation data of the display screen to be detected is parameter-marked according to the performance evaluation coefficient of the display screen to be detected, so as to obtain the performance evaluation data of the display screen to be detected.

Images