A Dynamic Detection Method for Pixel-Level Defects in Display Screens Based on Multispectral Imaging

Abstract

This invention relates to the field of display screen inspection technology, specifically to a dynamic detection method for pixel-level defects in display screens based on multispectral imaging. The method includes the following steps: employing multi-band parallel imaging to acquire high-resolution images of multiple fixed bands during dynamic display; performing filtered edge tracking and blurred region interpolation compensation, and using dynamic threshold updates to segment individual physical pixels in dynamic scenes; extracting three multi-dimensional features of pixels—multi-band light intensity, band correlation, and spatiotemporal dynamics—and using multi-band cross-validation to identify and classify pixel-level defects in the display screen, outputting the identification and classification data; through the above methods, dynamic adaptation and segmentation accuracy are improved, meeting the quality control requirements of large-scale production of high-resolution, high-refresh-rate display screens.

Description

Technical Field

[0001] This invention relates to the field of display screen inspection technology, and in particular to a dynamic detection method for pixel-level defects in display screens based on multispectral imaging. Background Technology

[0002] As a core interactive component of electronic devices, displays are widely used in smartphones, televisions, automotive displays, industrial control, and other fields. With the rapid iteration of display technology towards ultra-high resolution (8K and above), ultra-high refresh rates (120Hz and above), and tiny pixel sizes (≤5μm), the pixel density of various displays such as LCD, OLED, MiniLED, and MicroLED continues to increase. Pixel-level defects (such as single-pixel dead pixels, micro-spots, motion blur, color shift, and response delay) have an increasingly significant impact on display performance and product reliability. Accurate detection of pixel-level defects has become a core aspect of display manufacturing and quality control.

[0003] Existing methods for detecting display screen defects mainly include manual visual inspection, single-band machine vision inspection, traditional multispectral inspection, and electrical performance testing. However, they suffer from the following problems in pixel-level defect detection in dynamic scenes:

[0004] Poor adaptability to dynamic scenes: Traditional detection methods are mostly based on image acquisition in the static display state of the screen, which is difficult to simulate the dynamic refresh scene in actual use. As a result, dynamic defects such as dynamic ghosting, inter-frame color shift, and response delay cannot be effectively identified. Some dynamic detection solutions use single-band imaging technology and often achieve multi-band acquisition by switching filters. There is a time difference between bands (≥10ms), which cannot synchronously capture the instantaneous optical characteristics of pixels under dynamic display, resulting in the loss of defect information.

[0005] Insufficient pixel segmentation accuracy: Existing segmentation algorithms mostly employ fixed thresholds or static edge detection strategies, making it difficult to adapt to issues such as brightness fluctuations and motion blur during dynamic display processes. In dynamic scenes, fixed thresholds are prone to segmentation failure due to sudden brightness changes, while traditional edge detection algorithms are susceptible to image jitter and dynamic motion blur, resulting in edge offset and contour distortion. The single-pixel segmentation error is typically ≥0.2 pixels, which cannot meet the requirements for accurate localization of pixel-level defects.

[0006] Therefore, a dynamic detection method for pixel-level defects in displays that balances segmentation accuracy and dynamic scene adaptability is proposed to solve the problems of poor dynamic adaptation and low segmentation accuracy in existing technologies, and to meet the quality control requirements of large-scale production of high-resolution, high-refresh-rate displays. Summary of the Invention

[0007] The purpose of this invention is to provide a dynamic detection method for pixel-level defects in display screens based on multispectral imaging, aiming to solve the technical problems of poor dynamic adaptation and low segmentation accuracy in the prior art.

[0008] To achieve the above objectives, the present invention employs a dynamic detection method for pixel-level defects in a display screen based on multispectral imaging, comprising the following steps:

[0009] A multi-band parallel imaging method is used to acquire high-resolution images of multiple fixed bands while the display screen is dynamically showing the image.

[0010] Perform filtering edge tracking and blur region interpolation compensation, and use dynamic threshold update to segment individual physical pixels in dynamic scenes;

[0011] The system extracts three multi-dimensional features from pixels: multi-band light intensity, band correlation, and spatiotemporal dynamics. Multi-band cross-validation is used to identify and classify pixel-level defects in the display screen, and the identification and classification data are output.

[0012] In the step of acquiring high-resolution images of multiple fixed bands while dynamically displaying them on a screen using a multi-band parallel imaging method:

[0013] An imaging device was constructed consisting of a high frame rate CCD camera, a multi-band parallel filter assembly, a dynamic focus compensation unit, and an ambient light shield. The multi-band parallel filter assembly integrates a beam splitter prism and a filter array, including four fixed bands and three replaceable bands.

[0014] Lock the refresh rate of the imaging device and the display screen;

[0015] The imaging equipment is activated, and image data of each fixed band is acquired synchronously through the multi-band parallel filter assembly.

[0016] After the step of locking the refresh rate of the imaging device and the display screen:

[0017] The brightness change curve of the display screen is collected, and the focus parameters are adjusted in real time according to the position of the brightness peak to perform dynamic focus compensation.

[0018] In the steps of performing filtered edge tracking and blurred region interpolation compensation, and using dynamic threshold updates to segment individual physical pixels in dynamic scenes:

[0019] Set the update cycle, calculate the mean and variance of pixel grayscale within the window in real time, dynamically adjust the segmentation threshold, and adapt to the brightness fluctuations of the display screen.

[0020] Extract the initial edge of the pixel, predict the edge position of the next frame, and correct the edge offset in dynamic scenes;

[0021] Determine the grayscale change rate between image frames and perform interpolation compensation for blurred areas that exceed a preset threshold;

[0022] By combining the pixel registration results of multi-band images and using the pixel contour of the near-infrared band image with strong anti-interference as a benchmark, the segmentation deviation of the visible light band is corrected to segment individual physical pixels.

[0023] Among the steps, the following steps are involved: setting the update cycle, calculating the mean and variance of pixel grayscale within the window in real time, dynamically adjusting the segmentation threshold, and adapting to the brightness fluctuations of the display screen:

[0024] Based on the screen refresh rate and dynamic display switching speed, the update cycle of the sliding window is set to 5-10 frames.

[0025] For each band image acquired by multispectral acquisition, continuous frames are extracted at a set period to form a sliding window. All pixels within the window are traversed, and the gray value of each pixel is counted.

[0026] Based on statistical grayscale data, the mean and variance of pixel grayscale within the window are calculated to reflect the overall brightness level and grayscale distribution characteristics of the image within that period.

[0027] Based on the calculated mean and variance, the segmentation threshold of the Otsu's method is dynamically adjusted to match the current window's brightness.

[0028] In the steps of extracting the initial edge of a pixel, predicting the edge position of the next frame, and correcting edge offset in dynamic scenes:

[0029] The preprocessed multi-band image is processed, and reasonable high and low thresholds are set to filter out regions with abrupt changes in pixel grayscale, and the edge contours of individual pixels are initially extracted to form an initial edge dataset.

[0030] Based on the initial edge data, an edge motion trajectory model is constructed. Combined with the refresh rate of the display screen and the switching rules of the dynamic display, the direction and speed of edge motion are determined.

[0031] Based on the edge position information of the previous frame, predict the pixel edge positions in the current frame and the next frame;

[0032] The predicted edge positions are compared with the actual detected edge positions in the current frame, and the position deviation value is calculated.

[0033] The actual detected edge position is corrected based on the deviation value to compensate for the edge offset caused by dynamic display and imaging jitter, so that the edge contour always accurately fits the physical boundary of a single pixel.

[0034] In the step of determining the grayscale change rate between image frames and interpolating to compensate for blurred areas exceeding a preset threshold:

[0035] Select two consecutive frames of images from the same source band, and calculate the grayscale difference of corresponding pixels according to their pixel coordinates;

[0036] By comparing the grayscale difference with the time interval between two frames, the inter-frame grayscale change rate of each pixel is obtained, which quantifies the dynamic blur of the image.

[0037] A preset grayscale change rate threshold is set, and the grayscale change rate of each pixel is compared with the preset threshold. Pixel areas with change rates exceeding the threshold are marked as dynamic blur areas.

[0038] For the marked blurry area, find 3 to 5 adjacent clear images and extract the pixel contour and grayscale distribution information of the corresponding area in the clear frames;

[0039] Using the pixel outline of the clear frame as a reference, the pixel outline of the blurred area is supplemented and corrected to fill in the outline breaks and distortions caused by the blur.

[0040] Among them, in the step of combining the pixel registration results of multi-band images, using the pixel contour of the near-infrared band image with strong anti-interference as a benchmark, correcting the segmentation deviation of the visible light band, and segmenting individual physical pixels:

[0041] Pixel-level registration is performed on the multispectral images of each band, the spatial offset between different band images is calculated, and all band images are aligned to the same pixel coordinate system.

[0042] The pixel contours of the near-infrared band image are selected as the reference contours. The preliminary segmentation contours of the red, green, and blue visible light band images are extracted and compared with the reference contours of the near-infrared band one by one according to the pixel coordinates.

[0043] Calculate the positional deviation between the outline of each pixel in the visible light band and the reference outline;

[0044] Based on the calculated deviation value, the initial segmentation contour of the visible light band is corrected point by point to align the pixel contour of the visible light band with the reference contour.

[0045] By integrating the segmentation results of each band, the precise boundary of a single physical pixel is finally determined, and the single physical pixel is completely segmented.

[0046] Among them, in the steps of extracting three multi-dimensional features of pixels—multi-band light intensity, band correlation, and spatiotemporal dynamics—and using multi-band cross-validation to identify and classify pixel-level defects of the display screen, and outputting the identification and classification data:

[0047] For each segmented individual pixel unit, multi-band light intensity features such as peak value, valley value, and dynamic fluctuation coefficient of light intensity in each band are extracted, and light intensity penetration depth features in the near-infrared band are collected.

[0048] Calculate the band correlation characteristics of light intensity ratio, spectral similarity coefficient, and light intensity phase difference between bands to distinguish the differences in optical properties of different types;

[0049] Extract the spatiotemporal dynamic features of pixel light intensity rise and fall time constant, inter-frame light intensity change rate, and color switching response delay time in dynamic display sequences to capture defect features in dynamic scenes.

[0050] By integrating three types of feature parameters to construct a multi-dimensional feature vector, preliminary defect identification is performed through feature weighting and global correlation analysis to obtain suspected defect pixels.

[0051] After integrating the three types of feature parameters to construct a multi-dimensional feature vector, and performing preliminary defect identification through feature weighting and global correlation analysis to obtain suspected defect pixels:

[0052] Calculate the feature deviation rate of each band of the suspected defective pixel. If the preset conditions are met, it is determined to be a real defect; otherwise, a second acquisition and verification is triggered.

[0053] The system categorizes defect types, generates detection data containing defect location coordinates, type, feature deviation value, and confidence level, and outputs the data.

[0054] This invention discloses a dynamic detection method for pixel-level defects in display screens based on multispectral imaging. It employs multi-band parallel imaging to acquire high-resolution images of multiple fixed bands during dynamic display. The method performs filtering edge tracking and blurred region interpolation compensation, and uses dynamic threshold updates to segment individual physical pixels in dynamic scenes. It extracts three multi-dimensional features of pixels: multi-band light intensity, band correlation, and spatiotemporal dynamics. Multi-band cross-validation is then used to identify and classify pixel-level defects in the display screen, and the identification and classification data are output. Through these methods, dynamic adaptation and segmentation accuracy are improved, meeting the quality control requirements of large-scale production of high-resolution, high-refresh-rate display screens. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1This is a flowchart of the steps of the dynamic detection method for pixel-level defects in a display screen based on multispectral imaging according to the present invention.

[0057] Figure 2 This is a flowchart of steps S100 of the present invention.

[0058] Figure 3 This is a flowchart of steps S200 of the present invention.

[0059] Figure 4 This is a flowchart of steps S300 of the present invention.

[0060] Figure 5 This is a schematic diagram of the structural principle of the display screen pixel-level defect dynamic detection system based on multispectral imaging of the present invention.

[0061] Figure 6 This is a schematic diagram of the electronic device of the present invention.

[0062] 401 - Multispectral image acquisition module, 402 - Pixel-level dynamic segmentation module, 403 - Feature extraction and defect recognition module. Detailed Implementation

[0063] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0064] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0065] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0066] Please see Figures 1-4 This invention provides a dynamic detection method for pixel-level defects in a display screen based on multispectral imaging, comprising the following steps:

[0067] S100: Employs a multi-band parallel imaging method to acquire high-resolution images of multiple fixed bands while the display screen is dynamically showing the image.

[0068] In this embodiment, a multi-band parallel imaging method is employed to acquire high-resolution images of multiple fixed bands while the display screen is dynamically showing the image. The specific process is as follows:

[0069] S101: Construct an imaging device consisting of a high frame rate CCD camera, a multi-band parallel filter assembly, a dynamic focus compensation unit, and an ambient light shield; the multi-band parallel filter assembly integrates a beam splitter prism and a filter array, including four fixed bands and three replaceable bands.

[0070] S102: Lock the refresh rate of the imaging device and the display screen;

[0071] S103: Collects the brightness change curve of the display screen and adjusts the focus parameters in real time according to the brightness peak position to perform dynamic focus compensation;

[0072] S104: Start the imaging equipment and synchronously acquire image data of each fixed band through the multi-band parallel filter assembly.

[0073] In the above process, an imaging device consisting of a high frame rate CCD camera, a multi-band parallel filter assembly, a dynamic focus compensation unit, and an ambient light shield was constructed. The multi-band parallel filter assembly adopts an integrated design of a beam splitter prism and filter array, enabling simultaneous separation and acquisition of light signals from multiple bands. This assembly includes four fixed bands (red light 620~680nm, green light 520~580nm, blue light 440~480nm, and near-infrared 800~900nm) and three replaceable bands (supporting special band extensions such as ultraviolet and mid-infrared) to meet the detection needs of different types of displays. The high frame rate CCD camera has a frame rate of no less than 120fps to ensure dynamic acquisition of high refresh rate displays. The ambient light shield is made of light-shielding material, effectively isolating external light interference and controlling the ambient light intensity below 50 lux.

[0074] The high-precision clock synchronization module locks the refresh rate of the imaging device and the display screen, and adjusts the acquisition frame rate of the imaging device according to the actual refresh rate of the display screen (60Hz~240Hz) to ensure that the synchronization error between the two is ≤±0.5Hz. For high refresh rate displays with a refresh rate ≥120Hz, a frame rate doubling technology is adopted to make the acquisition frame rate of the imaging device twice the refresh rate of the display screen, ensuring that enough frames of images are acquired in each refresh cycle and avoiding information loss caused by dynamic ghosting.

[0075] The dynamic focus compensation unit starts working and collects the brightness change curve of the display screen in real time during the dynamic display process. By analyzing the brightness peak position in the curve, the clear imaging area of ​​the pixel is determined. Based on the brightness peak position, the focus parameters of the imaging device are adjusted in real time, with a focus response time of ≤5ms, to ensure the image clarity of a single pixel during the dynamic display process, so that the pixel spread function (PSF) is ≤1.2, which meets the image quality requirements of pixel-level detection.

[0076] After the imaging equipment parameters are calibrated, the imaging equipment is started, and the display screen operates according to the preset dynamic display scheme (including the switching sequence of different gray levels and color modes); the multi-band parallel filter component synchronously separates the light signals of each fixed band, and the high frame rate CCD camera synchronously acquires the image data of each band. The image resolution is set to 1.5 to 2 times the pixel density of the display screen to ensure that a single physical pixel can form a complete pixel point in each band image, so as to achieve comprehensive capture of multi-band optical information.

[0077] S200: Performs filtered edge tracking and blurred region interpolation compensation, and uses dynamic threshold updates to segment individual physical pixels in dynamic scenes.

[0078] In this embodiment, edge tracking and blurred region interpolation compensation are performed, and dynamic threshold updates are used to segment individual physical pixels in dynamic scenes. The specific process is as follows:

[0079] S201: Set the update cycle, calculate the mean and variance of pixel grayscale within the window in real time, dynamically adjust the segmentation threshold, and adapt to the brightness fluctuations of the display screen.

[0080] Based on the screen refresh rate and dynamic display switching speed, the update cycle of the sliding window is set to 5-10 frames.

[0081] For each band image acquired by multispectral acquisition, continuous frames are extracted at a set period to form a sliding window. All pixels within the window are traversed, and the gray value of each pixel is counted.

[0082] Based on statistical grayscale data, the mean and variance of pixel grayscale within the window are calculated to reflect the overall brightness level and grayscale distribution characteristics of the image within that period.

[0083] Based on the calculated mean and variance, the segmentation threshold of the Otsu's method is dynamically adjusted to match the current window's brightness.

[0084] S202: Extract the initial edge of the pixel, predict the edge position of the next frame, and correct the edge offset in dynamic scenes;

[0085] The preprocessed multi-band image is processed, and reasonable high and low thresholds are set to filter out regions with abrupt changes in pixel grayscale, and the edge contours of individual pixels are initially extracted to form an initial edge dataset.

[0086] Based on the initial edge data, an edge motion trajectory model is constructed. Combined with the refresh rate of the display screen and the switching rules of the dynamic display, the direction and speed of edge motion are determined.

[0087] Based on the edge position information of the previous frame, predict the pixel edge positions in the current frame and the next frame;

[0088] The predicted edge positions are compared with the actual detected edge positions in the current frame, and the position deviation value is calculated.

[0089] The actual detected edge position is corrected based on the deviation value to compensate for the edge offset caused by dynamic display and imaging jitter, so that the edge contour always accurately fits the physical boundary of a single pixel.

[0090] S203: Determine the grayscale change rate between image frames and perform interpolation compensation for blurred areas that exceed a preset threshold;

[0091] Select two consecutive frames of images from the same source band, and calculate the grayscale difference of corresponding pixels according to their pixel coordinates;

[0092] By comparing the grayscale difference with the time interval between two frames, the inter-frame grayscale change rate of each pixel is obtained, which quantifies the dynamic blur of the image.

[0093] A preset grayscale change rate threshold is set, and the grayscale change rate of each pixel is compared with the preset threshold. Pixel areas with change rates exceeding the threshold are marked as dynamic blur areas.

[0094] For the marked blurry area, find 3 to 5 adjacent clear images and extract the pixel contour and grayscale distribution information of the corresponding area in the clear frames;

[0095] Using the pixel outline of the clear frame as a reference, the pixel outline of the blurred area is supplemented and corrected to fill in the outline breaks and distortions caused by the blur.

[0096] S204: Combining the pixel registration results of multi-band images, and using the pixel contour of the near-infrared band image with strong anti-interference as a benchmark, the segmentation deviation of the visible light band is corrected to segment individual physical pixels.

[0097] Pixel-level registration is performed on the multispectral images of each band, the spatial offset between different band images is calculated, and all band images are aligned to the same pixel coordinate system.

[0098] The pixel contours of the near-infrared band image are selected as the reference contours. The preliminary segmentation contours of the red, green, and blue visible light band images are extracted and compared with the reference contours of the near-infrared band one by one according to the pixel coordinates.

[0099] Calculate the positional deviation between the outline of each pixel in the visible light band and the reference outline;

[0100] Based on the calculated deviation value, the initial segmentation contour of the visible light band is corrected point by point to align the pixel contour of the visible light band with the reference contour.

[0101] By integrating the segmentation results of each band, the precise boundary of a single physical pixel is finally determined, and the single physical pixel is completely segmented.

[0102] In the above process, an update cycle is set, and the mean and variance of pixel grayscale within the window are calculated in real time. The segmentation threshold is dynamically adjusted to adapt to the brightness fluctuations of the display screen. Specifically, based on the refresh rate of the display screen and the dynamic display switching speed, the update cycle of the sliding window is set to 5-10 frames to match the frequency of segmentation threshold adjustment with the frequency of brightness fluctuations. For each band of multispectral image acquisition, consecutive frames are extracted at a set cycle to form a sliding window. All pixels within the window are traversed, and the grayscale value of each pixel is counted. Based on the statistical grayscale data, the mean and variance of pixel grayscale within the window are calculated to reflect the overall brightness level and grayscale distribution characteristics of the image within that cycle. Based on the calculated mean and variance, the segmentation threshold of the Otsu's method (OTSU) is dynamically adjusted to ensure that the segmentation threshold always matches the brightness state of the current window, avoiding segmentation failure when the display screen brightness fluctuates, and ensuring the initial accuracy of pixel segmentation under different brightness scenarios.

[0103] The process involves extracting initial pixel edges, predicting edge positions for the next frame, and correcting edge offsets in dynamic scenes. Specifically, the Canny edge detection algorithm is used to process the pre-processed multi-band image. By setting reasonable high and low thresholds (the high threshold is 60%~70% of the maximum image grayscale value, and the low threshold is 30%~40% of the high threshold), regions with abrupt changes in pixel grayscale are selected, and the edge contours of individual pixels are initially extracted to form an initial edge dataset. Based on the initial edge data, combined with the refresh rate of the display screen and the switching rules of dynamic displays, an edge motion trajectory model is constructed to determine the possible motion direction and speed of the edges. Using the edge position information of the previous frame and the motion trajectory model, the theoretical positions of pixel edges in the current frame and the next frame are predicted. The predicted edge positions are compared with the actual detected edge positions in the current frame to calculate the position deviation value. A Kalman filter algorithm is introduced to correct the actual detected edge positions based on the deviation value, compensating for edge offsets caused by dynamic display screen and imaging jitter, ensuring that the edge contours always accurately fit the physical boundaries of individual pixels, and improving the stability of edge detection in dynamic scenes.

[0104] The method involves determining the inter-frame grayscale change rate of an image and performing interpolation compensation for blurred areas exceeding a preset threshold. Specifically, two consecutive frames of images from the same source band are selected, and the grayscale difference between corresponding pixels is calculated based on their pixel coordinates. The grayscale difference is compared with the time interval between the two frames to obtain the inter-frame grayscale change rate of each pixel, thereby quantifying the dynamic blurring degree of the image. A preset grayscale change rate threshold (ranging from 0.3 to 0.5) is used, determined based on the display refresh rate, imaging frame rate, and detection accuracy requirements. The entire image is traversed, and the grayscale change rate of each pixel is compared with the preset threshold to mark the pixel areas with change rates exceeding the threshold, i.e., the dynamic blurring areas. For the marked blurring areas, 3 to 5 adjacent clear images (frames with grayscale change rates below the threshold) are searched, and the pixel contours and grayscale distribution information of the corresponding areas in these clear frames are extracted. A bilinear interpolation algorithm is used, with the pixel contours of the clear frames as a reference, to supplement and correct the pixel contours of the blurring areas, filling in the contour breaks or distortions caused by blurring, and ensuring the integrity and accuracy of pixel segmentation within the blurring areas.

[0105] Combining pixel registration results from multi-band images, and using the pixel contours of near-infrared band images (which exhibit strong anti-interference properties) as a benchmark, the segmentation deviation of the visible light band is corrected to segment individual physical pixels. Specifically, the phase correlation method is used to perform pixel-level registration processing on multispectral images of each band (red, green, blue, near-infrared, etc.), calculating the spatial offsets (including translation and rotation deviations) between different band images, and aligning all band images to the same pixel coordinate system to ensure that the coordinate positions of the same physical pixel are consistent across different band images. Due to the strong resistance to ambient light interference and high penetration of the near-infrared band, the pixel contours of its images are less affected by external interference and are more stable. To ensure high accuracy, the pixel contours of near-infrared images were selected as the baseline contours. Preliminary segmentation contours were extracted from visible light band images (red, green, blue, etc.) after processing in previous steps and compared pixel by pixel with the baseline contours of the near-infrared band. The positional deviation between each pixel contour in the visible light band and the baseline contours was calculated. Based on the calculated deviation values, the preliminary segmentation contours in the visible light bands were corrected point by point to ensure full alignment between the pixel contours and the baseline contours. After correction, the segmentation results of each band were integrated to finally determine the precise boundary of each physical pixel, achieving complete segmentation of each physical pixel and providing accurate pixel units for subsequent feature extraction.

[0106] S300: Extracts three multi-dimensional features of pixels: multi-band light intensity, band correlation, and spatiotemporal dynamics. Employs multi-band cross-validation to identify and classify pixel-level defects on the display screen and outputs identification and classification data.

[0107] In this embodiment, three multi-dimensional features—multi-band light intensity, band correlation, and spatiotemporal dynamics—are extracted from pixels. Multi-band cross-validation is used to identify and classify pixel-level defects in the display screen, and the identification and classification data are output. The specific process is as follows:

[0108] S301: For a single pixel unit after segmentation, extract multi-band light intensity features such as peak value, valley value and dynamic fluctuation coefficient of light intensity in each band, and collect light intensity penetration depth features in the near-infrared band.

[0109] S302: Calculate the band correlation characteristics of light intensity ratio, spectral similarity coefficient, and light intensity phase difference between bands to distinguish the differences in optical properties of different types;

[0110] S303: Extracts the spatiotemporal dynamic features of pixel light intensity rise and fall time constant, inter-frame light intensity change rate, and color switching response delay time in dynamic display sequences, and captures defect features in dynamic scenes.

[0111] S304: Integrate three types of feature parameters to construct a multi-dimensional feature vector, and perform preliminary defect identification through feature weighting and global correlation analysis to obtain suspected defect pixels;

[0112] S305: Calculate the feature deviation rate of each band of the suspected defective pixel. If the preset conditions are met, it is determined to be a real defect; otherwise, a second acquisition and verification is triggered.

[0113] S306: Classify the defect types, generate detection data containing defect location coordinates, type, feature deviation value, and confidence level, and output the data.

[0114] In the above process, for each individual pixel unit segmented in step S200, multi-band light intensity features such as peak light intensity, valley light intensity, and dynamic fluctuation coefficient of light intensity for each fixed band (red light, green light, blue light, near infrared) are extracted one by one. The dynamic fluctuation coefficient of light intensity is obtained by calculating the ratio of the standard deviation to the mean of the light intensity value of the pixel in the dynamic display sequence. At the same time, the light intensity penetration depth feature of the near infrared band is collected. This feature is obtained by analyzing the attenuation degree of near infrared light in the pixel area and is used to detect hidden defects in the internal circuit of the display screen.

[0115] Based on the extracted multi-band light intensity features, the light intensity ratios between each band (such as the ratio of red light to green light intensity, and the ratio of near-infrared light to blue light intensity), the spectral similarity coefficient based on the spectral angle matching algorithm, and the band correlation features such as the light intensity phase difference between bands are calculated. By distinguishing the differences in optical characteristics between normal pixels and defective pixels through the differences in these feature parameters, for example, color shift defects will be manifested by a spectral similarity coefficient that is significantly lower than the normal range, and bright spot defects will be manifested by an abnormally high light intensity ratio in a specific band.

[0116] Extract the spatiotemporal dynamic features of pixels in the dynamic display sequence, including the light intensity rise time constant, the light intensity fall time constant (obtained by fitting the light intensity change curve), the inter-frame light intensity change rate (the ratio of the light intensity difference between consecutive frames to the time interval), and the color switching response delay time (the time required for the pixel light intensity to stabilize when switching from one color mode to another). These features are used to capture the unique defect features in dynamic scenes. For example, the image retention defect is manifested as a continuously low inter-frame light intensity change rate, and the response delay defect is manifested as a color switching response delay time exceeding the normal range.

[0117] By integrating three types of feature parameters—multi-band light intensity, band correlation, and spatiotemporal dynamics—a multi-dimensional feature vector with more than 18 dimensions is constructed. The feature vector is then input into a deep learning model that integrates an attention mechanism and a Transformer structure. The model uses the attention mechanism to weight and enhance defect-related features and suppress irrelevant background interference. Finally, a Transformer encoder is used to perform global correlation analysis of multi-band and multi-frame features, complete the preliminary defect identification, and output the suspected defect pixels and their corresponding confidence scores (the confidence score threshold is set to 0.8).

[0118] A normal pixel feature benchmark library is constructed (containing 100,000+ normal pixel feature data of different types of displays). For the suspected defective pixels obtained in step S304, the normal pixel feature parameters of the corresponding type of display in the benchmark library are queried, and the deviation rate of each band feature of the suspected defective pixel from the benchmark value is calculated. Preset deviation rate judgment conditions (minor defects ≥5%, moderate defects ≥10%, severe defects ≥20%) are set. When the feature deviation rate of at least 3 bands meets the preset conditions of the corresponding defect level, it is judged as a real defect. If the above conditions are not met, the imaging device is triggered to perform a second acquisition of the area (focused magnification 2 times), and the feature extraction and deviation rate calculation are re-executed for verification to avoid missed detection.

[0119] Based on the output of the deep learning model and the conclusions of multi-band cross-validation, real defects are classified into eight types, including bright spots, dark spots, color casts, dead pixels, afterimages, response delays, uneven segmentation, and latent defects. Detection data is generated, including defect location coordinates (based on the display screen pixel coordinate system), defect type, deviation values ​​of each feature parameter, defect severity level, and confidence level. The detection data is output to the production line MES system via an Ethernet interface, and a visual detection report is generated, supporting intuitive display such as defect image screenshots and feature curve comparisons, providing a basis for subsequent quality traceability and rectification.

[0120] Corresponding to the aforementioned embodiments of the dynamic detection method for pixel-level defects in displays based on multispectral imaging, this application also provides embodiments of a dynamic detection system for pixel-level defects in displays based on multispectral imaging.

[0121] Figure 5 This is a block diagram illustrating a dynamic detection system for pixel-level defects in a display screen based on multispectral imaging, according to an exemplary embodiment. (Refer to...) Figure 5 The system may include: a multispectral image acquisition module 401, a pixel-level dynamic segmentation module 402, and a feature extraction and defect recognition module 403; wherein:

[0122] The multispectral image acquisition module 401 is used to acquire high-resolution images of multiple fixed bands while the display screen is dynamically displayed, using a multi-band parallel imaging method.

[0123] The pixel-level dynamic segmentation module 402 is used to perform filtered edge tracking and blurred region interpolation compensation, and uses dynamic threshold update to segment individual physical pixels in dynamic scenes.

[0124] The feature extraction and defect identification module 403 is used to extract three types of multi-dimensional features of pixels: multi-band light intensity, band correlation, and spatiotemporal dynamics. It uses multi-band cross-validation to identify and classify pixel-level defects of the display screen and outputs identification and classification data.

[0125] In this embodiment, the multispectral image acquisition module 401 adopts a multi-band parallel imaging method to acquire high-resolution images of multiple fixed bands during dynamic display. The pixel-level dynamic segmentation module 402 performs filtering edge tracking and blurred region interpolation compensation, and uses dynamic threshold updates to segment individual physical pixels in dynamic scenes. The feature extraction and defect recognition module 403 extracts three types of multi-dimensional features of pixels: multi-band light intensity, band correlation, and spatiotemporal dynamics. It uses multi-band cross-validation to identify and classify pixel-level defects of the display screen and outputs identification and classification data. Through the above methods, dynamic adaptation and segmentation accuracy are improved, meeting the quality control requirements of large-scale production of high-resolution, high-refresh-rate displays.

[0126] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0127] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0128] Accordingly, this application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the above-described method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging. Figure 6 The diagram shown is a hardware structure diagram of any device with data processing capabilities, which is part of a dynamic detection system for pixel-level defects in a display screen based on multispectral imaging, according to an embodiment of the present invention. Except for... Figure 6 In addition to the processor, memory, and network interface shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0129] Accordingly, this application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the aforementioned method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging. The computer-readable storage medium can be an internal storage unit of any data-processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data-processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data-processing device, and can also be used to temporarily store data that has been output or will be output.

[0130] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0131] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A dynamic detection method for pixel-level defects in a display screen based on multispectral imaging, characterized in that, Includes the following steps: A multi-band parallel imaging method is used to acquire high-resolution images of multiple fixed bands while the display screen is dynamically showing the image. Filtered edge tracking and blurred region interpolation compensation are performed, and dynamic threshold updates are used to segment individual physical pixels in dynamic scenes; the specific process is as follows: The update cycle is set, and the mean and variance of pixel grayscale within the window are calculated in real time. The segmentation threshold is dynamically adjusted to adapt to the brightness fluctuations of the display screen. Based on the refresh rate and dynamic display switching speed of the display screen, the update cycle of the sliding window is set to 5-10 frames. For each band of multispectral image acquisition, consecutive frames are extracted at the set cycle to form a sliding window. All pixels within the window are traversed, and the grayscale value of each pixel is counted. Based on the statistical grayscale data, the mean and variance of pixel grayscale within the window are calculated to reflect the overall brightness level and grayscale distribution characteristics of the image within that cycle. Based on the calculated mean and variance, the segmentation threshold of the Otsu's method is dynamically adjusted to make the segmentation threshold fit the brightness state of the current window. Extract the initial edge of the pixel, predict the edge position of the next frame, and correct the edge offset in dynamic scenes; The preprocessed multi-band image is processed by setting reasonable high and low thresholds to filter out regions with abrupt changes in pixel grayscale, and the edge contours of individual pixels are initially extracted to form an initial edge dataset. Based on the initial edge data, an edge motion trajectory model is constructed, and the edge motion direction and speed are determined by combining the refresh rate of the display screen and the switching rules of dynamic display. According to the edge position information of the previous frame, the pixel edge position in the current frame and the next frame is predicted. The predicted edge position is compared with the actual detected edge position in the current frame to calculate the position deviation value. The actual detected edge position is corrected according to the deviation value to compensate for the edge offset caused by dynamic display and imaging jitter, so that the edge contour always accurately fits the physical boundary of the individual pixel. Determine the grayscale change rate between image frames and perform interpolation compensation for blurred areas that exceed a preset threshold; By combining the pixel registration results of multi-band images and using the pixel contour of the near-infrared band image with strong anti-interference as a benchmark, the segmentation deviation of the visible light band is corrected and individual physical pixels are segmented. The system extracts three multi-dimensional features from pixels: multi-band light intensity, band correlation, and spatiotemporal dynamics. Multi-band cross-validation is used to identify and classify pixel-level defects in the display screen, and the identification and classification data are output.

2. The method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging as described in claim 1, characterized in that, In the step of acquiring high-resolution images of multiple fixed bands while dynamically displaying them on a screen using a multi-band parallel imaging method: An imaging device was constructed consisting of a high frame rate CCD camera, a multi-band parallel filter assembly, a dynamic focus compensation unit, and an ambient light shield. The multi-band parallel filter assembly integrates a beam splitter prism and a filter array, including four fixed bands and three replaceable bands. Lock the refresh rate of the imaging device and the display screen; The imaging equipment is activated, and image data of each fixed band is acquired synchronously through the multi-band parallel filter assembly.

3. The method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging as described in claim 2, characterized in that, After the steps of locking the refresh rate of the imaging device and the display screen: The brightness change curve of the display screen is collected, and the focus parameters are adjusted in real time according to the position of the brightness peak to perform dynamic focus compensation.

4. The method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging as described in claim 1, characterized in that, In the step of determining the grayscale change rate between image frames and interpolating to compensate for blurred areas exceeding a preset threshold: Select two consecutive frames of images from the same source band, and calculate the grayscale difference of corresponding pixels according to their pixel coordinates; By comparing the grayscale difference with the time interval between two frames, the inter-frame grayscale change rate of each pixel is obtained, which quantifies the dynamic blur of the image. A preset grayscale change rate threshold is set, and the grayscale change rate of each pixel is compared with the preset threshold. Pixel areas with change rates exceeding the threshold are marked as dynamic blur areas. For the marked blurry area, find 3 to 5 adjacent clear images and extract the pixel contour and grayscale distribution information of the corresponding area in the clear frames; Using the pixel outline of the clear frame as a reference, the pixel outline of the blurred area is supplemented and corrected to fill in the outline breaks and distortions caused by the blur.

5. The method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging as described in claim 1, characterized in that, In the step of segmenting individual physical pixels by combining pixel registration results from multi-band images and using the pixel contours of near-infrared band images with strong anti-interference capabilities as a benchmark, and correcting segmentation deviations in the visible light band: Pixel-level registration is performed on the multispectral images of each band, the spatial offset between different band images is calculated, and all band images are aligned to the same pixel coordinate system. The pixel contours of the near-infrared band image are selected as the reference contours. The preliminary segmentation contours of the red, green, and blue visible light band images are extracted and compared with the reference contours of the near-infrared band one by one according to the pixel coordinates. Calculate the positional deviation between the outline of each pixel in the visible light band and the reference outline; Based on the calculated deviation value, the initial segmentation contour of the visible light band is corrected point by point to align the pixel contour of the visible light band with the reference contour. By integrating the segmentation results of each band, the precise boundary of a single physical pixel is finally determined, and the single physical pixel is completely segmented.

6. The method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging as described in claim 1, characterized in that, In the steps of extracting three multi-dimensional features of pixels—multi-band light intensity, band correlation, and spatiotemporal dynamics—and using multi-band cross-validation to identify and classify pixel-level defects on the display screen, and outputting the identification and classification data: For each segmented individual pixel unit, multi-band light intensity features such as peak value, valley value, and dynamic fluctuation coefficient of light intensity in each band are extracted, and light intensity penetration depth features in the near-infrared band are collected. Calculate the band correlation characteristics of light intensity ratio, spectral similarity coefficient, and light intensity phase difference between bands to distinguish the differences in optical properties of different types; Extract the spatiotemporal dynamic features of pixel light intensity rise and fall time constant, inter-frame light intensity change rate, and color switching response delay time in dynamic display sequences to capture defect features in dynamic scenes. The three types of feature parameters are integrated to construct a multi-dimensional feature vector. Preliminary defect identification is performed through feature weighting and global correlation analysis to obtain suspected defect pixels.

7. The method for dynamic detection of pixel-level defects in a display screen based on multispectral imaging as described in claim 6, characterized in that, After integrating the three types of feature parameters to construct a multi-dimensional feature vector, and performing preliminary defect identification through feature weighting and global correlation analysis to obtain suspected defect pixels: Calculate the feature deviation rate of each band of the suspected defective pixel. If the preset conditions are met, it is determined to be a real defect; otherwise, a second acquisition and verification is triggered. The system categorizes defect types, generates detection data containing defect location coordinates, type, feature deviation value, and confidence level, and outputs the data.

Images