Method and system for detecting and positioning faults of narrow-frame TFT-LCD (Thin Film Transistor-Liquid Crystal Display) display unit

Abstract

The invention provides a method and a system for detecting and positioning faults of a narrow-frame TFT-LCD (Thin Film Transistor-Liquid Crystal Display) display unit. Belongs to the technical field of display. The method comprises the following steps: carrying out image acquisition on a narrow-frame TFT-LCD display unit through a high-resolution camera; carrying out the preprocessing of the collected image; based on an image processing technology, fault features existing in a display unit are extracted, pixels in an image are subjected to point-by-point analysis by setting a specific algorithm, abnormal pixel points or areas are identified, and feature information is extracted. Through a high-resolution camera and a fine image processing technology, detail information of the display unit can be captured, and accurate extraction of a fault area is ensured.

Description

Technical Field

[0001] The invention provides a narrow-frame TFT-LCD display unit fault detection and positioning method and system, belonging to the field of display technology. Background Art

[0002] With the continuous advancement of display technology, narrow-border TFT-LCD display units have been widely used in electronic devices such as smartphones, tablets, and TVs due to their advantages of lightness, high definition, and low power consumption. However, the space limitations brought about by the narrow-border design make the fault detection and location of display units more complicated. Traditional detection methods often rely on manual visual inspection or the use of expensive professional equipment, which is not only inefficient but also difficult to accurately identify subtle fault points. Therefore, it is particularly important to develop an efficient and accurate method for fault detection and location of narrow-border TFT-LCD display units. Summary of the invention

[0003] The present invention provides a method and system for detecting and locating faults in a narrow-frame TFT-LCD display unit, which are used to solve the problems mentioned in the above background technology: The present invention provides a method for detecting and locating a fault of a narrow-frame TFT-LCD display unit, the method comprising: S1, collecting images of the narrow-frame TFT-LCD display unit through a high-resolution camera; and preprocessing the collected images; S2. Extract the fault features in the display unit based on image processing technology, analyze the pixels in the image point by point by setting a specific algorithm, identify abnormal pixels or areas, and extract feature information; S3. Based on the machine learning algorithm, a fault classification and location model is constructed, and a large number of display unit images with known faults are used as training samples to train the model; during the training process, the parameters of the model are continuously optimized; S4. Send a specific test signal to the display unit and observe its response to determine whether there is a fault; and use high-precision measuring instruments to accurately measure the fault point; S5. According to the results of image analysis and hardware detection, determine the specific location of the fault point and give corresponding repair suggestions.

[0004] The present invention provides a narrow-frame TFT-LCD display unit fault detection and positioning system, the system comprising: Data acquisition module: collects images of the narrow-frame TFT-LCD display unit through a high-resolution camera and pre-processes the collected images; Feature extraction module: Based on image processing technology, the fault features existing in the display unit are extracted. By setting a specific algorithm, the pixels in the image are analyzed point by point, abnormal pixels or areas are identified, and feature information is extracted.

[0005] Model training module: Based on the machine learning algorithm, a fault classification and location model is constructed, and a large number of display unit images with known faults are used as training samples to train the model; during the training process, the model parameters are continuously optimized; Fault judgment module: by sending a specific test signal to the display unit and observing its response, it can be judged whether there is a fault; and the fault point can be accurately measured using high-precision measuring instruments; Suggestion feedback module: Determine the specific location of the fault point based on the results of image analysis and hardware detection, and give corresponding repair suggestions.

[0006] The beneficial effects of the present invention are as follows: through high-resolution cameras and sophisticated image processing technology, detailed information of the display unit can be captured to ensure accurate extraction of the fault area; by combining image processing with machine learning, hidden fault features can be extracted from a large amount of data, and the false detection rate can be reduced through model optimization; by combining multiple detection methods such as image analysis, hardware testing, and dynamic analysis, a comprehensive analysis of faults can be performed from different angles and levels, thereby effectively avoiding missed detection and misdiagnosis; periodic faults such as dynamic flickering can be tracked through time series analysis technology, further improving the accuracy of detection; the system can automatically identify potential fault areas based on image acquisition, and automatically generate fault reports based on electrical test results, greatly improving the accuracy of fault diagnosis and prevention. The automation level of the detection process is improved; after the fault is located, a specific repair plan is provided in combination with in-depth analysis to ensure the efficiency and pertinence of fault handling; during the training process, data enhancement, regularization and other technologies are used to effectively avoid the overfitting problem and ensure that the model has strong generalization ability; through continuous iterative optimization, high-performance fault detection and classification capabilities can be achieved, and the accuracy and reliability of fault detection can be improved; through accurate fault location and classification, unnecessary manual troubleshooting and repeated testing can be reduced, detection efficiency can be improved, and a lot of time and cost can be saved; during the repair process, detailed fault analysis and cost evaluation can ensure the economy and feasibility of the repair plan and avoid unnecessary waste of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a step diagram of the method of the present invention; Figure 2 This is a system module diagram of the present invention. DETAILED DESCRIPTION

[0008] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0009] One embodiment of the present invention, as Figure 1 As shown, a method for detecting and locating faults in a narrow-frame TFT-LCD display unit is provided, the method comprising: S1, collecting images of the narrow-frame TFT-LCD display unit through a high-resolution camera; and preprocessing the collected images; S2. Based on image processing technology, the fault features existing in the display unit are extracted, and the features include line defects, point defects, uneven display, flickering, and poor grayscale. By setting a specific algorithm, the pixels in the image are analyzed point by point to identify abnormal pixels or areas, and extract their position, size, shape and other feature information.

[0010] S3. Based on the machine learning algorithm, a fault classification and location model is constructed, and a large number of display unit images with known faults are used as training samples to train the model. During the training process, the parameters of the model are continuously optimized. After the training is completed, the model can automatically identify and locate the fault points in the new display unit images. S4. Verify the image analysis results through hardware detection technology, send specific test signals to the display unit, observe its response, and determine whether there is a fault; and use high-precision measuring instruments to accurately measure the fault point; S5. According to the results of image analysis and hardware detection, determine the specific location of the fault point and give corresponding repair suggestions.

[0011] The working principle of the above technical solution is as follows: use a high-resolution camera to shoot the narrow-frame TFT-LCD display unit to ensure that the captured image is clear enough for subsequent analysis; automatically adjust the camera exposure time according to the current ambient light to avoid overexposure or underexposure of the image and ensure that the image brightness is moderate; adjust the image color so that it can remain natural under different lighting conditions, eliminate color cast, and improve the color accuracy of the image; perform denoising, contrast enhancement and other processing on the collected image to improve the image quality, laying a good foundation for subsequent feature extraction and fault identification; use image processing technology to identify fault features such as line defects (such as broken lines, uneven colors), point defects (such as bright spots, dark spots), uneven display (such as local brightness differences), flickering (such as screen jitter), and poor grayscale (such as blurred color levels) in the display unit; set a specific algorithm to analyze each pixel in the image one by one, and identify abnormal pixels or areas by comparing the difference between pixel values ​​(such as brightness, chromaticity) and normal values; extract the position (such as coordinates) and size (such as area) of abnormal pixels or areas ), shape (such as circle, line) and other feature information to provide data support for subsequent fault classification and location; select appropriate machine learning algorithms (such as convolutional neural network CNN, support vector machine SVM, etc.) to build a fault classification and location model; collect a large number of display unit images with known faults as training samples to ensure the diversity and representativeness of the samples; use the training samples to train the model, continuously adjust the model parameters, and improve the recognition accuracy and generalization ability of the model; after the training is completed, apply the model to new display unit images to automatically identify and locate the fault point; send specific test signals to the display unit and observe its response, such as brightness changes, color display, etc., to determine whether there is a fault; use high-precision measuring instruments (such as microscopes, spectrometers, etc.) to accurately measure the fault point and obtain more detailed fault information; compare the hardware detection results with the image analysis results to verify the accuracy of the image analysis, and supplement and improve the fault information; determine the specific location of the fault point based on the results of image analysis and hardware detection; provide different repair plans for different types of faults. Such as replacing damaged components (such as LED lamp beads, driver chips, etc.), adjusting circuit parameters (such as voltage, current, etc.), cleaning the display screen, etc.; according to the repair suggestions, repair the display unit to restore its normal display function.

[0012] The effects of the above technical solution are as follows: through the high-resolution camera combined with adaptive exposure and automatic white balance technology, clear images can be obtained under different lighting conditions, providing high-quality input for subsequent fault feature extraction; based on image processing technology, the fault features such as line defects, point defects, display unevenness, flicker, and poor grayscale in the display unit are accurately extracted, reducing the error of human judgment and improving the accuracy of fault detection; by building and training a machine learning model, the fault point in the new display unit image can be automatically identified and located, realizing the automation and intelligence of fault detection and improving the detection efficiency; by sending a specific test signal to the display unit, observing its response, and combining high-precision measuring instruments to accurately measure the fault point, further The reliability of the image analysis results was verified in the first step, reducing the risk of false alarms and missed alarms; the fault point was accurately located by combining the results of image analysis and hardware detection, providing accurate information support for subsequent repair work; according to the specific location and type of the fault point, corresponding repair suggestions were given to avoid unnecessary repair operations and reduce repair costs; by extracting and analyzing fault characteristics, potential faults can be discovered in time, providing a scientific basis for fault prevention and maintenance, and extending the service life of the display unit; this technical solution can identify and locate various types of faults and has a wide range of applicability; with the continuous advancement of image processing technology and machine learning algorithms, this technical solution can be easily upgraded and optimized to adapt to new fault types and detection requirements.

[0013] In one embodiment of the present invention, the S1 includes: S11, using a high-resolution camera to perform all-round image acquisition of the narrow-frame TFT-LCD display unit to capture every detail of the display unit; S12, based on adaptive exposure technology, automatically adjusts the exposure parameters of the camera according to the ambient light, captures images under different lighting conditions, and applies automatic white balance technology to eliminate color deviation; S13, preprocessing the collected image, wherein the preprocessing includes denoising, contrast enhancement, and edge sharpening.

[0014] The working principle of the above technical solution is: a high-resolution camera is used to perform omnidirectional image acquisition of the narrow-frame TFT-LCD display unit. This means that the camera will shoot the display unit from multiple angles or positions to ensure that every detail of the display unit is captured. Omnidirectional acquisition helps to more accurately identify and analyze the fault characteristics in the subsequent steps; the technology monitors the changes in ambient light in real time through the built-in light sensor, and automatically adjusts the camera's exposure parameters (such as shutter speed, aperture size, etc.) according to these changes. In this way, the camera can capture images with moderate brightness and clear details regardless of whether it is in strong light or weak light environment; by analyzing the color information in the image, the color balance of the camera is automatically adjusted to eliminate color deviation. It ensures that the colors in the image remain natural and accurate under different lighting conditions. This helps to more accurately identify and analyze the color information of the fault characteristics in the subsequent steps; the image may be interfered by various noises during the acquisition process, such as electronic noise, optical noise, etc. The denoising step reduces or eliminates these noises by applying algorithms such as filters to make the image clearer and smoother; contrast refers to the brightness difference between the bright and dark parts of the image. The contrast enhancement step adjusts the brightness distribution of the image to make the bright parts brighter and the dark parts darker, thereby increasing the visual layering of the image, which helps to more clearly identify the fault features in the subsequent steps; the edge is the dividing line between objects in the image and is also one of the important manifestations of the fault features. The edge sharpening step enhances the contrast of the edges in the image to make the edges clearer and more prominent, which helps to more accurately identify and locate the fault features in the subsequent steps.

[0015] The effects of the above technical solution are as follows: by using a high-resolution camera to perform all-round image acquisition of the narrow-frame TFT-LCD display unit, every detail of the display unit can be captured. This is crucial for accurately identifying and analyzing fault features in subsequent steps; the exposure parameters of the camera are automatically adjusted according to the ambient light, so that images with moderate brightness and clear details can be captured under different lighting conditions. This avoids the problem of overexposure or underexposure of images caused by excessive or weak light, and improves the flexibility and adaptability of image acquisition; the application of automatic white balance technology can eliminate color deviation and ensure that the colors in the image remain natural and accurate. This helps to more accurately identify and analyze the color information of fault features in subsequent steps; through denoising, noise interference in the image can be reduced or eliminated, making the image clearer and smoother. This helps to improve the accuracy and reliability of fault feature identification in subsequent steps; contrast enhancement processing can increase the visual layering of the image and make the brightness difference between the bright and dark parts more obvious. This helps to more clearly identify fault features in subsequent steps and improve the accuracy of detection; sharp edge processing can enhance the contrast of the edges in the image, making the edges clearer and more prominent. This is of great significance for accurately identifying and locating fault features in subsequent steps; through high-resolution cameras, adaptive exposure technology, automatic white balance technology, and image preprocessing steps, an automated process for image acquisition and preprocessing is achieved. This greatly improves detection efficiency and reduces manual intervention and errors; high-quality image acquisition and preprocessing provide strong support for accurately identifying and analyzing fault features in subsequent steps. This helps to promptly discover and deal with potential problems in the display unit and improve the accuracy and reliability of detection.

[0016] In one embodiment of the present invention, the S2 includes: S21, refine the fault characteristics of the TFT-LCD display unit; and after image preprocessing, perform preliminary screening through image statistical characteristics (such as mean, variance, histogram, etc.) to exclude images that are obviously normal or too abnormal (such as completely black screen, completely white screen); S22. For each pixel, calculate the difference between it and the surrounding pixels, and determine whether there is an abnormality based on a preset threshold; and record the position of the abnormal pixel, the difference value and the statistical characteristics of the surrounding pixels; the difference is obtained by the following formula: ; Among them, there is a pixel point P with a gray value of , the grayscale values ​​of its surrounding pixels (for example, the four neighbors above, below, left and right) are Represents pixel P and its The absolute value of the grayscale difference of the neighboring pixels. The summation symbol represents the sum of the differences of the four neighboring pixels, and then divided by 4 to get the average difference.

[0017] S23, performing neighborhood analysis on the initially identified abnormal pixels to determine whether they constitute a specific fault mode (such as continuous abnormal pixels forming lines, clustered abnormal pixels forming spots, etc.); further confirming the existence of the fault by calculating the number, density, morphology and other characteristics of the abnormal pixels in the neighborhood; S24, extracting detailed feature information from the confirmed fault area, and performing comparison and classification using a preset fault feature library based on the extracted feature information; S25. Further process the fault area using morphological operations (such as dilation, erosion, opening operation, closing operation, etc.), and perform context analysis based on multiple factors to identify the fault mode that depends on the specific background or environment based on the analysis results; S26. Summarize all the extracted fault features and their classification results to form a fault feature list or a fault distribution map, and verify the extracted fault features using known fault samples or manual inspection.

[0018] The working principle of the above technical solution is as follows: the fault characteristics of the TFT-LCD display unit are defined in detail, and in addition to the basic fault types, more detailed fault types are also considered. Each fault type has clear visual characteristics and possible physical causes. Then, after image preprocessing, the image is preliminarily screened using the statistical characteristics of the image (such as mean, variance, histogram, etc.) to exclude images that are obviously normal or too abnormal. For example, if the brightness mean of an image is very low and the variance is very small, it may mean that the image is completely black and should be excluded; for each pixel, the difference between it and the surrounding pixels is calculated, and whether the pixel is abnormal is determined based on a preset threshold. If the difference of a pixel exceeds the threshold, the pixel is considered abnormal, and its position, difference value and statistical characteristics of the surrounding pixels are recorded. This step helps to preliminarily identify pixels that may have faults; the neighborhood analysis of the initially identified abnormal pixels is performed to determine whether they constitute a specific fault mode. The existence of the fault is further confirmed by calculating the number, density, morphology and other characteristics of the abnormal pixels in the neighborhood. For example, if continuous abnormal pixels form lines, it may indicate the presence of line defects; if clustered abnormal pixels form spots, it may indicate the presence of point defects; for confirmed fault areas, detailed feature information is extracted, including shape, size, color, texture, and dynamic characteristics. Then, the preset fault feature library is used for comparison and classification. The fault feature library contains typical feature descriptions and threshold ranges for various types of faults, so that the fault area can be accurately classified into specific fault types. For faults that cannot be directly classified, they are marked as unknown types and left for subsequent analysis; the fault area is further processed using morphological operations (such as dilation, corrosion, opening operation, closing operation, etc.) to improve its shape and edge features. Then, context analysis is performed based on multiple factors, including the position of the fault area in the entire image, its relative relationship with other areas, and the overall layout of the image. This step helps to identify fault modes that depend on specific backgrounds or environments, such as edge faults or faults within specific patterns; all extracted fault features and their classification results are summarized to form a fault feature list or fault distribution map. Then, the extracted fault features are verified using known fault samples or manual inspection to ensure their accuracy and reliability. This step is an important part of the fault detection process and helps ensure the accuracy and credibility of the final test results.

[0019] The effect of the above technical solution is: by refining the fault characteristics of the TFT-LCD display unit, not only the basic fault types are covered, but also more detailed fault types are considered. This helps to more accurately identify various faults during the detection process and improve the precision and accuracy of the detection; after image preprocessing, the image statistical characteristics (such as mean, variance, histogram, etc.) are used for preliminary screening to exclude obviously normal or too abnormal images, such as completely black screens, all white screens, etc. This helps to reduce the false detection rate and improve the detection efficiency; by calculating the difference between each pixel and the surrounding pixels, and judging whether there is an abnormality based on the preset threshold, the abnormal pixel point can be accurately located. This helps to further analyze and process the fault area; recording the position, difference value and statistical characteristics of the abnormal pixel and the surrounding pixels provides rich data support for subsequent analysis. This helps to have a deeper understanding of the characteristics and causes of the fault; by performing neighborhood analysis on the initially identified abnormal pixels, it is determined whether it constitutes a specific fault mode, such as continuous abnormal pixels forming lines, clustered abnormal pixels forming spots, etc. This helps to accurately determine the existence and type of faults; by calculating the number, density, morphology and other features of abnormal pixels in the neighborhood, the existence of faults can be further confirmed. This helps to enhance the reliability and stability of detection; for confirmed fault areas, detailed feature information is extracted, including but not limited to shape, size, color, texture and dynamic characteristics. This provides comprehensive data support for subsequent comparison and classification; using the preset fault feature library for comparison and classification, the fault area can be accurately classified into a specific fault type. This helps to quickly locate the fault and take corresponding repair measures; further processing the fault area using morphological operations can improve its shape and edge features, providing more favorable data for subsequent analysis and processing; context analysis based on multiple factors can identify fault modes that depend on specific backgrounds or environments, such as edge faults, faults within specific patterns, etc. This helps to gain a deeper understanding of the causes and characteristics of the faults; all extracted fault features and their classification results are summarized to form a fault feature list or a fault distribution map. This helps to intuitively understand the distribution and characteristics of the faults; verifying the extracted fault features using known fault samples or manual inspections can ensure the accuracy and credibility of the detection results. This helps to provide strong support for subsequent repair work. The above formula calculates the average grayscale difference between the central pixel and its neighboring pixels, and it can be judged whether the pixel is abnormal. Greater than the preset threshold T , then the pixel P is considered abnormal.

[0020] In one embodiment of the present invention, the S25 includes: S251, performing morphological processing on the fault area, and performing refinement processing on the fault area based on the morphological processing; S252, analyzing the relative position relationship between the fault area and the edge, center or other key areas of the image based on the position of the fault area in the entire image; S253, analyzing the relative relationship between the fault area and other normal or abnormal areas, and analyzing whether the fault area is located in a specific pattern, icon or text area in combination with the overall layout of the image; S254, extracting features of the fault area using a deep learning network, dividing the image into multiple areas using an image segmentation algorithm, and performing fault detection on each area; S255. For dynamic flicker faults, the time series analysis technology is combined to capture the changing rules of the faults, and the extracted fault features are verified using known fault samples, and the wrong or omitted fault features are corrected and supplemented; S256. Feedback the verification results to the previous processing steps to optimize algorithm parameters, improve processing strategies, and enhance feature extraction capabilities.

[0021] The working principle of the above technical solution is as follows: morphological processing is performed on the fault area, which includes basic operations such as dilation, erosion, opening and closing operations. The dilation operation can expand the boundary of the fault area and connect the broken parts; the erosion operation can shrink the boundary of the fault area and separate the overlapping parts. The opening operation (erosion first and then dilation) helps to remove small noise points, while the closing operation (dilation first and then erosion) helps to fill small holes. These operations work together on the fault area to eliminate noise, optimize shape and edge features; based on the morphological processing, the fault area is further refined, such as skeleton extraction and contour tracking. Skeleton extraction can reveal the internal structure of the fault area, such as the width of the line and the core position of the spot, while contour tracking can accurately depict the boundary of the fault area. These refinement processes provide more detailed information, which is helpful for subsequent feature extraction and classification; by analyzing the position of the fault area in the entire image, its relative position relationship with the edge, center or other key areas of the image is determined. This helps to identify fault modes that may depend on specific locations, such as edge faults or center faults; analyze the relative relationship between the fault area and other normal or abnormal areas, including distance, direction, overlap, etc. At the same time, combined with the overall layout of the image, it is determined whether the fault area is located in a specific pattern, icon or text area. This helps to identify fault modes that may depend on a specific environment or background; the deep learning network is used to extract features from the fault area, which can capture more complex fault features. At the same time, the image segmentation algorithm is used to divide the image into multiple regions, and fault detection is performed separately. This helps to locate the fault area more accurately and extract richer fault features; for dynamic flickering faults, the time series analysis technology is combined to capture the changing laws of the fault. This includes analyzing the time, frequency, duration and other characteristics of the fault. Then, the extracted fault features are verified using known fault samples to ensure the accuracy and reliability of the features. For erroneous or omitted fault features, corrections and supplements are made to improve the accuracy and accuracy of detection; the verification results are fed back to previous processing steps, such as morphological processing, context analysis, etc. According to the verification results, the algorithm parameters are optimized, the processing strategy is improved, and the feature extraction capability is enhanced. This forms a closed-loop feedback mechanism that helps to continuously improve the accuracy and efficiency of fault detection.

[0022] The effect of the above technical solution is: through morphological processing, such as expansion, corrosion, opening and closing operations, noise can be eliminated, broken parts can be connected, overlapping parts can be separated, and the edges of the fault area can be smoothed. This helps to more accurately extract the morphological features of the fault area and provide a more reliable basis for subsequent feature extraction and classification; based on the morphological processing, refinement processing, such as skeleton extraction and contour tracking, can further reveal the internal structure of the fault area, such as the width of the line, the core position of the spot, etc. This information is of great significance for in-depth understanding of fault characteristics and improving classification accuracy. By analyzing the position of the fault area in the entire image and its relative position relationship with the edge, center or other key areas of the image, the fault area can be located more accurately, reducing the possibility of misjudgment and missed judgment. In combination with the overall layout of the image, analyzing whether the fault area is located in a specific pattern, icon or text area helps to identify fault modes that may depend on a specific environment or background, and improve the pertinence and accuracy of detection. Using a deep learning network to extract features from the fault area can capture more complex fault features and improve the accuracy and comprehensiveness of feature extraction. By dividing the image into multiple areas through an image segmentation algorithm and performing fault detection separately, refined detection and analysis of the fault area can be achieved, improving the accuracy and efficiency of detection. For dynamic flickering faults, combining time series analysis technology to capture the changing laws of the fault can help to deeply understand the dynamic characteristics of the fault and improve the sensitivity and accuracy of detection. Using known fault samples to verify the extracted fault features can ensure the accuracy and reliability of the features. Correcting and supplementing the fault features that are wrong or missed can help improve the accuracy and completeness of detection; feeding back the verification results to the previous processing steps, such as morphological processing, context analysis, etc., can form a closed-loop feedback system. This helps to continuously optimize algorithm parameters, improve processing strategies, and enhance feature extraction capabilities to improve overall detection performance; through continuous feedback and optimization, the detection algorithm and process can be continuously improved, the accuracy, efficiency and reliability of fault detection can be improved, and more powerful support can be provided for fault detection and repair of TFT-LCD display units.

[0023] In one embodiment of the present invention, the S254 includes: According to the characteristics of the fault area, a deep learning model is selected, and feature extraction is performed based on the selected deep learning model; Collect fault image samples, annotate them, identify the location and type of the fault area, and train the deep learning model to learn the feature representation of the fault area; According to the complexity of the fault area and the overall layout of the image, the image segmentation algorithm is selected and the segmentation results are post-processed; Use the trained deep learning model to extract features from the segmented fault area, and perform feature selection and dimensionality reduction operations; The extracted feature vector is input into the classifier to detect the fault area, and the detected fault area is further classified; According to the feature vector of the fault area and the known fault type library, a classifier is used to determine the specific type of the fault; The detection results are verified using known fault samples and fed back to the deep learning model, image segmentation algorithm, and feature extraction and classification steps.

[0024] The working principle of the above technical solution is as follows: First, according to the characteristics of the fault area (such as shape, size, color, texture, etc.), a suitable deep learning model is selected. These models may include convolutional neural networks (CNN), residual networks (ResNet), VGG, etc. The specific selection depends on the complexity of the fault characteristics and the performance of the model; then, the deep learning model is trained using the collected fault image samples. These samples should include various fault types and faults of different severity, and have been labeled to clarify the location and type of the fault area. Through training, the model can learn the characteristic representation of the fault area; according to the complexity of the fault area and the overall layout of the image, a suitable image segmentation algorithm is selected. These algorithms may include threshold-based segmentation, edge detection segmentation, region growing segmentation, clustering segmentation, and deep learning-based segmentation algorithms (such as MaskR-CNN, FCN, etc.); the segmentation results are post-processed, such as removing small noise areas, filling holes, smoothing edges, etc., to optimize the segmentation effect; the trained deep learning model is used to extract features from the segmented fault area. These features may include shape features (such as perimeter, area, aspect ratio, etc.), color features (such as RGB value, HSV value, brightness, saturation, etc.), texture features (such as gray-level co-occurrence matrix, local binary pattern, etc.) and high-level semantic features (such as features automatically learned by deep learning models); then feature selection and dimensionality reduction operations are performed to reduce redundant features and improve the performance and efficiency of the classifier; the extracted feature vector is input into the classifier, such as support vector machine (SVM), random forest, K nearest neighbor (KNN) or deep learning classifier (such as fully connected neural network, convolutional neural network, etc.); the classifier detects the fault area and further classifies the detected fault area to determine the specific type of fault; the detection results are verified using known fault samples to ensure the accuracy of classification; the verification results are fed back to the deep learning model, image segmentation algorithm, and feature extraction and classification steps. According to the verification results, the model parameters are optimized, the segmentation strategy is improved, the feature extraction capability is enhanced, and the classification accuracy is improved.

[0025] The effect of the above technical solution is: the selection of deep learning models according to the characteristics of the fault area can ensure the matching degree between the model and the fault features, thereby improving the accuracy and efficiency of feature extraction. For example, for fault areas with complex textures, a convolutional neural network (CNN) or its variants with strong feature extraction capabilities can be selected; fault image samples containing various fault types and different degrees of severity are collected and accurately labeled, which can train a deep learning model with stronger generalization ability. This model can learn the feature representation of the fault area and has high accuracy for the detection and classification of unknown faults; according to the complexity of the fault area and the overall layout of the image, the selection of a suitable image segmentation algorithm can ensure the accurate segmentation of the fault area. At the same time, post-processing of the segmentation results, such as removing small noise areas, filling holes, smoothing edges, etc., can further optimize the segmentation effect and improve the accuracy of subsequent feature extraction and classification; using the trained deep learning model for feature extraction can capture a variety of features of the fault area, including shape, color, texture, and high-level semantic features. These features can provide rich information for the classifier and improve the accuracy of classification. At the same time, feature selection and dimensionality reduction operations can reduce redundant features and improve the efficiency and performance of the classifier; inputting the extracted feature vector into the classifier can achieve accurate detection of the fault area. For the detected fault area, further classification operations can determine the specific type of fault. This classification has high accuracy and can provide strong support for subsequent fault repair and maintenance; using known fault samples to verify the detection results, and feeding the verification results back to the deep learning model, image segmentation algorithm, and feature extraction and classification steps, can form a closed-loop feedback system. This feedback mechanism can continuously optimize model parameters, improve segmentation strategies, enhance feature extraction capabilities, and improve classification accuracy, thereby achieving continuous improvement and enhancement.

[0026] In one embodiment of the present invention, the S255 includes: Capture the occurrence process of dynamic flicker faults through high-speed camera equipment or video recording technology, and pre-process the captured dynamic fault data; The periodic characteristics of dynamic flicker faults are analyzed by Fourier transform. By identifying the frequency and period of fault occurrence, the correlation between the fault and the external environment or internal factors of the system is obtained. Use trend line fitting to analyze the long-term trend of dynamic flicker faults; combine statistical methods and machine learning algorithms to detect abnormal points or mutation points in dynamic flicker faults; Based on the results of time series analysis, the key features of dynamic flicker faults are extracted, and the extracted fault features are verified using known fault samples; By comparing the matching degree between the actual fault features and the preset feature library, the accuracy and reliability of feature extraction are evaluated, and the features that do not meet expectations are corrected and supplemented; the accuracy and reliability are evaluated by the following formula: ; in, are weight coefficients, representing the relative importance of accuracy, classification accuracy, matching degree, and stability, respectively, and ; TP represents true positive examples (correctly identified fault features), TN represents true negative examples (correctly identified normal features), FP represents false positive examples (incorrectly identified fault features), and FN represents false negative examples (incorrectly identified normal features); Indicates The number of samples correctly classified into the fault category, n represents the total number of fault categories; Indicates The number of fault features that match the preset feature library, Indicates The total number of fault features; σ represents the standard deviation of multiple feature extraction results, and μ represents the average value of multiple feature extraction results; According to the verification results, the feature extraction algorithm is optimized, and the verified fault characteristics are fed back to the time series analysis step and the feature extraction algorithm. According to the feedback results, the time series analysis algorithm and the feature extraction algorithm are iteratively optimized.

[0027] The working principle of the above technical solution is: using high-speed camera equipment or video recording technology to capture the occurrence process of dynamic flicker faults in equipment or systems in real time; preprocessing the captured dynamic fault data, including removing noise (such as background noise, electromagnetic interference, etc.), enhancing contrast (to improve the visibility of fault characteristics), stabilizing frame rate (to ensure data continuity and stability), etc.; according to the requirements of the analysis algorithm, the preprocessed data is cropped or scaled to meet the needs of subsequent analysis steps; using Fourier transform to perform frequency domain analysis on the preprocessed dynamic flicker fault data; by identifying the frequency and period of the fault, the correlation between the fault and the external environment (such as power supply fluctuations, temperature changes, etc.) or internal factors of the system (such as component aging, design defects, etc.) is obtained; this helps to determine the root cause of the fault and provide guidance for subsequent fault repair and prevention; using trend line fitting technology to analyze the long-term change trend of dynamic flicker faults; combining statistical methods and machine learning algorithms to detect abnormal points or mutation points in dynamic flicker faults, which may represent critical moments or turning points in the development of the fault; by analyzing these abnormal points , we can further understand the development law and potential risks of faults; based on the results of time series analysis, extract the key features of dynamic flicker faults, such as flicker frequency, cycle length, change trend (such as rising, falling, stable, etc.), abnormal point location, etc.; these features can fully reflect the characteristics of the fault and provide a basis for subsequent fault classification, identification and prediction; use known fault samples to verify the extracted fault features to ensure the accuracy and reliability of the features; for features that do not meet expectations, correct and supplement them to improve the quality of feature extraction; according to the verification results, optimize the feature extraction algorithm, including adjusting algorithm parameters, introducing new feature extraction methods or improving feature representation methods, etc.; by continuously optimizing the feature extraction algorithm, the accuracy and efficiency of fault detection can be continuously improved; the verified fault features are fed back to the time series analysis step and feature extraction algorithm; according to the feedback results, the time series analysis algorithm and feature extraction algorithm are iteratively optimized to further improve the accuracy and efficiency of fault detection; this iterative optimization process can form a closed-loop feedback system to continuously promote the development and progress of fault detection technology.

[0028] The effects of the above technical solutions are as follows: through high-speed camera equipment or video recording technology, the occurrence process of dynamic flicker faults can be captured in real time and accurately; preprocessing steps (such as noise removal, contrast enhancement, frame rate stabilization, etc.) can significantly improve the clarity and accuracy of the data, providing a reliable basis for subsequent analysis; using Fourier transform to analyze the periodic characteristics of dynamic flicker faults can accurately identify the frequency and period of fault occurrence; it is helpful to establish the correlation between the fault and the external environment or internal factors of the system, so as to more accurately locate the cause of the fault; based on the results of time series analysis, the key features of dynamic flicker faults are extracted, such as flicker frequency, cycle length, change trend, abnormal point location, etc.; using known fault samples to verify the extracted fault features can ensure the accuracy and reliability of the features; using trend line fitting technology to analyze the long-term change trend of dynamic flicker faults can quickly identify the development trend of the fault; combining statistical methods and machine learning algorithms can efficiently detect abnormal points or Mutation points provide a basis for fault warning and timely processing; optimize the feature extraction algorithm according to the verification results, including adjusting algorithm parameters, introducing new feature extraction methods or improving feature representation methods; by continuously optimizing the feature extraction algorithm, the accuracy and efficiency of fault detection can be continuously improved, and false alarms and missed alarms can be reduced; the verified fault features are fed back to the time series analysis step and feature extraction algorithm, and the algorithm is iteratively optimized according to the feedback results; this iterative optimization process can form a closed-loop feedback system, continuously promoting the improvement and upgrading of fault detection technology; by continuously optimizing the algorithm and extracting more accurate fault features, the robustness of the system can be enhanced, so that it can remain stable and accurate in the face of complex and changeable fault conditions; accurate fault detection and analysis can help maintenance personnel quickly locate the fault point, reduce troubleshooting time, and thus reduce maintenance costs; by analyzing the long-term change trend and abnormal points of the fault, potential safety hazards can be discovered in time, and measures can be taken to prevent them, thereby improving the safety of the system. The above evaluation formula integrates multiple performance indicators, including accuracy (E), classification accuracy, matching degree and stability. The combination of these indicators can fully reflect the performance of the algorithm in different dimensions and avoid the deviation that may be caused by a single evaluation standard; true positives and true negatives can reflect the algorithm's ability to correctly identify fault features and normal features, ensuring that the model's judgment does not have too many errors; false positives and false negatives reflect the situation of algorithm misclassification, further helping to analyze the problems of the algorithm in the classification process. By optimizing these two indicators, the misjudgment rate can be reduced and the reliability of the algorithm can be improved; the recognition effect of different types of fault categories is considered through weighted classification accuracy to avoid the misclassification of certain categories affecting the overall evaluation results. This can help optimize classifiers for different fault types and improve overall diagnostic accuracy; by comparing the matching degree of the extracted fault features with the preset feature library, the quality of feature extraction can be evaluated.If the feature extraction results of a certain type of fault have a low match with the preset feature library, it means that the accuracy of feature extraction is poor and the feature extraction algorithm needs to be further adjusted. This helps to improve the feature extraction process and make it more consistent with the actual fault characteristics; stability is measured by calculating the ratio of the standard deviation of the feature extraction results to the mean value. A lower standard deviation indicates a higher consistency in the feature extraction process and better stability of the model; if the fluctuations in multiple extraction results are large, it means that the algorithm has poor stability, which may lead to inconsistent fault features and affect the reliability of diagnosis. Therefore, this indicator helps to discover instability in the algorithm and optimize it accordingly; the weighting coefficient allows the evaluation to adjust the relative importance of different indicators according to the actual application scenario. For example, if stability is more important than accuracy in a specific application scenario, it can be increased. The weight of the input feature is the weight of the output feature, and vice versa. This flexibility allows the formula to adapt to different fault diagnosis needs; by evaluating the results of the formula, the feature extraction algorithm and time series analysis method can be iteratively optimized. This feedback-based optimization mechanism helps to improve the model performance in continuous training and evaluation, thereby achieving more efficient and accurate fault detection and diagnosis; comprehensively considering accuracy, classification ability, feature matching degree and stability, the optimization goal of the entire fault detection system is more comprehensive, avoiding the situation where the optimization of a single dimension leads to performance degradation in other aspects, and achieving the effect of global performance improvement. The advantage of the above evaluation formula is that it not only considers the traditional accuracy indicators, but also adds multi-dimensional factors such as classification accuracy, feature matching and stability, making the optimization of the fault feature extraction algorithm more comprehensive and able to be flexibly adjusted according to different application requirements. This method helps to improve the performance of the dynamic flicker fault detection system, making it more stable and reliable in practical applications, and better adapting to changes in different types of faults.

[0029] In one embodiment of the present invention, S3 includes: S31, obtaining a large amount of image data of narrow-frame TFT-LCD display units with known faults, the data including images of various fault types (such as line defects, point defects, uneven display, flickering, poor grayscale, etc.) and normal display units, and annotating the collected image data in detail, including information such as fault type, location, severity, etc.; S32, dividing the labeled data set into a training set, a validation set and a test set, wherein the training set is used for model training, the validation set is used for model tuning and parameter selection, and the test set is used to evaluate the final performance of the model; S33. Select a machine learning model based on the characteristics of the fault detection and classification tasks; S34. Perform preliminary training on the model using the training set. Based on the results of the preliminary training, adjust and optimize the model's hyperparameters and use regularization techniques (such as L1 and L2 regularization) to prevent the model from overfitting. At the same time, evaluate the generalization ability of the model by monitoring the performance of the validation set and stop training early if necessary. S35. During the training process, the input data is continuously enhanced and the performance of the model is evaluated. Finally, the model is evaluated through the test set. S36. Based on the evaluation results of the test set, further optimize and improve the model, retrain the optimized model, and repeat the above evaluation and verification process until the model performance meets the requirements.

[0030] The working principle of the above technical solution is as follows: a large amount of image data of narrow-frame TFT-LCD display units with known faults is obtained from various channels; the data set includes various fault types, such as line defects, point defects, uneven display, flickering, poor grayscale, etc., as well as images of normal display units; the collected image data is annotated in detail, including information such as fault type, location, severity, etc., to provide accurate targets for subsequent model training; the annotated data set is divided into a training set, a validation set, and a test set; the training set is used for preliminary training of the model so that the model can learn the characteristics of the data; the validation set is used for model tuning and parameter selection, and the generalization ability of the model is evaluated by monitoring the performance of the validation set, and the hyperparameters of the model are adjusted; the test set is used to evaluate the final performance of the model to verify the generalization ability of the model and the actual application effect; according to the task characteristics of fault detection and classification, a suitable machine learning model is selected; possible models include convolutional neural networks (CNN), support vector machines (SVM), and random forests; Use the training set to conduct preliminary training on the model; adjust and optimize the model's hyperparameters based on the results of the preliminary training; use regularization techniques (such as L1 and L2 regularization) to prevent the model from overfitting and improve the model's generalization ability; evaluate the model's generalization ability by monitoring the performance of the validation set, and stop training in advance if necessary to avoid the model from overfitting on the training set; during the training process, continuously enhance the input data, such as rotating, scaling, flipping, etc., to increase the diversity of the data and the robustness of the model; evaluate the model's performance, including indicators such as accuracy, recall rate, and F1 score; use the test set to conduct a final evaluation of the model to verify the model's generalization ability and actual application effect; further optimize and improve the model based on the results of the test set evaluation; possible optimization measures include adjusting the model architecture, adding data enhancement strategies, using more complex loss functions, etc.; retrain the optimized model and repeat the above evaluation and verification process until the model performance meets the requirements.

[0031] The effects of the above technical solution are as follows: by obtaining a large amount of image data of narrow-border TFT-LCD display units with known faults, and annotating the fault type, location, severity and other information in detail, a rich and accurate data basis is provided for model training; a rich data set helps the model learn more fault features and improve the accuracy of fault detection; according to the task characteristics of fault detection and classification, a suitable machine learning model is selected, and preliminary training is performed through the training set; by monitoring the performance of the validation set, the hyperparameters of the model are adjusted and optimized to ensure that the model can accurately identify various fault types; regularization techniques (such as L1 and L2 regularization) are used to prevent model overfitting and improve the generalization ability of the model; by using strategies such as early stopping of training, the model is prevented from overfitting on the training set to ensure the accuracy of the model in practical applications; the labeled data set is divided into training set, validation set and test set to ensure that the model can be exposed to different data during training, tuning and evaluation; this data set division method helps to improve the generalization ability of the model , so that it can still maintain good performance when facing unknown data; during the training process, the input data is continuously enhanced, such as rotation, scaling, flipping, etc., to increase the diversity of data and the robustness of the model; data enhancement helps the model learn more feature changes and improve the model's adaptability to complex scenarios; according to the results of the test set evaluation, the model is further optimized and improved, including adjusting the model architecture, adding data enhancement strategies, using more complex loss functions, etc.; by continuously optimizing the model, the performance of the model is improved, so that it can more accurately identify the fault type and reduce false positives and false negatives; the optimized model is retrained, and the evaluation and verification process is repeated until the model performance meets the requirements; this iterative training and verification method helps to gradually approach the optimal solution and improve the performance and efficiency of the model; through the trained model, the automatic fault detection of the narrow-frame TFT-LCD display unit can be realized, reducing manual intervention and inspection costs; automatic fault detection helps to improve production efficiency and reduce production costs. ; The model can detect and warn potential faults in real time, providing timely information support for maintenance personnel; through early warning and preventive measures, the failure rate can be reduced and the service life of the equipment can be extended.

[0032] In one embodiment of the present invention, the S4 includes: S41, the test platform automatically starts the image acquisition device to acquire high-resolution images of the narrow-frame TFT-LCD display unit; and automatically transmits the acquired images to the image processing system. S42, the image acquisition system analyzes the image based on a pre-trained machine learning model to identify potential fault areas and types; S43. According to the result of the image analysis, the test platform automatically selects the corresponding electrical test instrument to perform an electrical performance test on the display unit, including the measurement of parameters such as voltage, current, resistance, and capacitance; S44. Combining the results of image analysis and electrical performance testing, the test platform automatically locates the fault area and type in the display unit; and automatically generates a detailed fault report, including the fault type, location, severity, possible causes, and recommended repair measures.

[0033] The working principle of the above technical solution is as follows: after receiving the test instruction or trigger signal, the test platform automatically starts the image acquisition device. This device is usually a high-resolution camera or image sensor, which is specially used to capture high-definition images of narrow-frame TFT-LCD display units; the collected image data is automatically transmitted to the image processing system. This process can be completed by wired (such as USB, Ethernet) or wireless (such as Wi-Fi, Bluetooth), depending on the hardware configuration and communication requirements of the test platform; the image data received by the image processing system is input into a pre-trained machine learning model. This model is trained on a large number of narrow-frame TFT-LCD display unit image data with known faults, and can identify potential fault areas and types in the image; the model extracts features (such as color, texture, shape, etc.) from the image and compares them with the fault features in the training data to identify potential fault areas and types; according to the results of the image analysis, the test platform automatically selects the corresponding electrical test instrument. These instruments may include voltmeters, ammeters, resistance meters, capacitor testers, etc., which are used to measure various electrical performance parameters of the display unit; the selected electrical test instrument performs performance testing on the display unit, and measures and records the values ​​of parameters such as voltage, current, resistance, and capacitance. These parameters help to further confirm the existence and nature of the fault; combined with the results of image analysis and electrical performance testing, the test platform automatically locates the fault area and type in the display unit. This process may involve comprehensive analysis and comparison of multiple test results; based on the results of fault location, the test platform automatically generates a detailed fault report. The report includes the fault type, location, severity, possible causes, and recommended repair measures. This information is presented in the form of electronic documents for subsequent analysis, processing, and archiving.

[0034] The effects of the above technical solution are as follows: through high-resolution image acquisition equipment, tiny fault details in the display unit can be captured, improving the accuracy of fault detection; the automated acquisition and transmission process reduces errors in human operation and improves detection efficiency; the pre-trained machine learning model can quickly and accurately analyze images and identify potential fault areas and types; the application of the model greatly shortens the time of fault detection and improves overall detection efficiency; by combining the results of image analysis and electrical performance testing, the fault area and type can be more accurately located; the electrical performance test provides additional fault information, which helps to further confirm the nature and severity of the fault; the generated fault report contains the fault type, location, Detailed information such as severity, possible causes, and recommended repair measures; this helps maintenance personnel quickly understand the fault situation and develop effective repair plans; the entire detection process is automated, reducing manual intervention and inspection costs; automated detection improves production efficiency and shortens product testing cycles; by timely discovering and locating faults, preventive measures can be taken to avoid further deterioration of the fault or impact on product quality; this helps reduce downtime and production cost losses caused by faults; as detection data accumulates, statistical analysis of fault types and distribution can be performed, providing strong support for product improvement and quality control; data analysis can also help identify potential problems in the production process and take measures in advance to make improvements.

[0035] In one embodiment of the present invention, S5 includes: S51, collecting detailed fault phenomenon descriptions from the aforementioned detection steps, conducting in-depth analysis on the collected fault data, and identifying possible causes of the faults; S52. Classify the fault into different types according to the fault phenomenon and data, where the different types include hardware fault, software fault and design defect; S53. Based on the fault classification and fault phenomenon, preliminarily determine which subsystem or component the fault may occur in, and conduct a detailed investigation of the preliminarily located subsystem or component using a test instrument; S54, gradually narrow down the possible scope of the fault by step-by-step elimination method until the specific fault point is located. After the fault point is located, verify whether the fault exists by simulating the fault condition or replacing the faulty component; S55. In combination with the results of fault location and the design principle of the display unit, the cause of the fault is analyzed in depth, and the accuracy and reliability of the cause of the fault are confirmed through further testing and verification; S56. Develop a specific repair plan based on the cause of the failure and conduct a cost assessment on the repair plan; S57. If the repair cost is too high or the repair difficulty is great, consider whether there are alternative solutions, such as using a spare display unit or upgrading the display unit, and repair the display unit according to the repair plan; S58. After the repair is completed, the display unit is subjected to a function test and a performance test, and the repair process, repair plan and repair result are recorded.

[0036] The working principle of the above technical solution is: collect detailed fault phenomenon descriptions from the previous detection steps (such as S4), including the time, location, environmental conditions and working status of the display unit when the fault occurred; conduct in-depth analysis of the collected fault data, including checking fault logs, analyzing detection data and images, etc., to identify possible causes of the fault; classify the faults into different types such as hardware faults, software faults and design defects according to the fault phenomena and data; the basis for classification may include the location where the fault occurs, the characteristics of the fault manifestation, the scope of the fault impact, etc.; based on the fault classification and fault phenomenon, preliminarily determine which subsystem or component the fault may occur in, such as a power module, a display module, a control circuit, etc.; use oscilloscopes, logic analyzers, signal generators and other test instruments to conduct detailed inspections of the preliminarily located subsystems or components. At the same time, use microscopes, X-ray detection and other equipment to inspect the physical structure to obtain more accurate fault information; gradually narrow the scope of possible faults through step-by-step elimination until the specific fault point is located; after locating the fault point, verify whether the fault exists by simulating fault conditions or replacing faulty components; combine the results of fault location and the design principle of the display unit to deeply analyze the cause of the fault, including physical damage, circuit failure, software error, design defect, etc.; confirm the accuracy and reliability of the cause of the fault through further testing and verification; formulate specific repair plans according to the cause of the fault, including replacing faulty components, repairing circuits, updating software, etc.; conduct cost assessment on the repair plan, including material cost, labor cost, time cost, etc., to ensure the economy and feasibility of the repair plan; if the repair cost is too high or the repair is difficult, consider whether there are alternatives, such as using spare display units, upgrading display units, etc.; repair the display unit according to the repair plan, including replacing faulty components, adjusting circuit parameters, updating software, etc.; after the repair is completed, perform functional and performance tests on the display unit to ensure that the fault has been completely repaired and the display unit can work normally; record the repair process, repair plan and repair results for subsequent analysis and improvement.

[0037] The effects of the above technical solution are: by collecting detailed descriptions of fault phenomena and deeply analyzing fault data (including fault logs, detection data, image analysis, etc.), the possible causes of the fault can be more accurately identified; by adopting a step-by-step elimination method, combined with detailed troubleshooting of test instruments and equipment, the scope of the fault can be gradually narrowed until the specific fault point is located. This method not only improves the accuracy of fault location, but also greatly improves efficiency; the faults are divided into different types such as hardware faults, software faults and design defects, which helps to analyze the faults more comprehensively and systematically; based on fault location and cause analysis, a specific repair plan is formulated, including replacing faulty components, repairing circuits, updating software, etc. The repair plan is targeted and can directly solve the fault problem; when formulating the repair plan, the material cost, labor cost, time cost, etc. are evaluated, which helps to find the best balance between cost control and repair effect; if the repair cost is too high or the repair is difficult, consider using alternative solutions such as spare display units or upgrading display units to ensure the continuity of production or service; after the repair is completed, the display unit is functionally tested and performance tested to ensure that the fault has been completely repaired and the display unit can work normally. This helps to ensure the quality of repair and avoid secondary failures. Recording the repair process, repair plan and repair results will help with subsequent analysis and improvement, and improve the level of fault handling. Technical Solution S5 has formed a set of standardized fault handling processes, which helps to improve overall maintenance efficiency and service quality. By recording and analyzing the fault handling process, valuable experience and knowledge can be accumulated, providing reference and reference for the subsequent handling of similar faults.

[0038] One embodiment of the present invention, as Figure 2 As shown, a narrow-frame TFT-LCD display unit fault detection and positioning system, the system comprises: Data acquisition module: collects images of the narrow-frame TFT-LCD display unit through a high-resolution camera and pre-processes the collected images; Feature extraction module: Based on image processing technology, the fault features in the display unit are extracted. By setting a specific algorithm, the pixels in the image are analyzed point by point, abnormal pixels or areas are identified, and feature information such as their position, size, and shape are extracted.

[0039] Model training module: Based on machine learning algorithms, a fault classification and location model is constructed, and a large number of display unit images with known faults are used as training samples to train the model. During the training process, the model parameters are continuously optimized. After the training is completed, the model can automatically identify and locate the fault points in new display unit images. Fault judgment module: Verify the image analysis results through hardware detection technology, send specific test signals to the display unit, observe its response, and judge whether there is a fault; and use high-precision measuring instruments to accurately measure the fault point; Suggestion feedback module: Determine the specific location of the fault point based on the results of image analysis and hardware detection, and give corresponding repair suggestions. For different types of faults, provide different repair solutions, such as replacing damaged components, adjusting circuit parameters, etc.

[0040] The working principle of the above technical solution is as follows: use a high-resolution camera to shoot the narrow-frame TFT-LCD display unit to ensure that the captured image is clear enough for subsequent analysis; automatically adjust the camera exposure time according to the current ambient light to avoid overexposure or underexposure of the image and ensure that the image brightness is moderate; adjust the image color so that it can remain natural under different lighting conditions, eliminate color cast, and improve the color accuracy of the image; perform denoising, contrast enhancement and other processing on the collected image to improve the image quality, laying a good foundation for subsequent feature extraction and fault identification; use image processing technology to identify fault features such as line defects (such as broken lines, uneven colors), point defects (such as bright spots, dark spots), uneven display (such as local brightness differences), flickering (such as screen jitter), and poor grayscale (such as blurred color levels) in the display unit; set a specific algorithm to analyze each pixel in the image one by one, and identify abnormal pixels or areas by comparing the difference between pixel values ​​(such as brightness, chromaticity) and normal values; extract the position (such as coordinates) and size (such as area) of abnormal pixels or areas ), shape (such as circle, line) and other feature information to provide data support for subsequent fault classification and location; select appropriate machine learning algorithms (such as convolutional neural network CNN, support vector machine SVM, etc.) to build a fault classification and location model; collect a large number of display unit images with known faults as training samples to ensure the diversity and representativeness of the samples; use the training samples to train the model, continuously adjust the model parameters, and improve the recognition accuracy and generalization ability of the model; after the training is completed, apply the model to new display unit images to automatically identify and locate the fault point; send specific test signals to the display unit and observe its response, such as brightness changes, color display, etc., to determine whether there is a fault; use high-precision measuring instruments (such as microscopes, spectrometers, etc.) to accurately measure the fault point and obtain more detailed fault information; compare the hardware detection results with the image analysis results to verify the accuracy of the image analysis, and supplement and improve the fault information; determine the specific location of the fault point based on the results of image analysis and hardware detection; provide different repair plans for different types of faults. Such as replacing damaged components (such as LED lamp beads, driver chips, etc.), adjusting circuit parameters (such as voltage, current, etc.), cleaning the display screen, etc.; according to the repair suggestions, repair the display unit to restore its normal display function.

[0041] The effects of the above technical solution are as follows: through the high-resolution camera combined with adaptive exposure and automatic white balance technology, clear images can be obtained under different lighting conditions, providing high-quality input for subsequent fault feature extraction; based on image processing technology, the fault features such as line defects, point defects, display unevenness, flicker, and poor grayscale in the display unit are accurately extracted, reducing the error of human judgment and improving the accuracy of fault detection; by building and training a machine learning model, the fault point in the new display unit image can be automatically identified and located, realizing the automation and intelligence of fault detection and improving the detection efficiency; by sending a specific test signal to the display unit, observing its response, and combining high-precision measuring instruments to accurately measure the fault point, further The reliability of the image analysis results was verified in the first step, reducing the risk of false alarms and missed alarms; the fault point was accurately located by combining the results of image analysis and hardware detection, providing accurate information support for subsequent repair work; according to the specific location and type of the fault point, corresponding repair suggestions were given to avoid unnecessary repair operations and reduce repair costs; by extracting and analyzing fault characteristics, potential faults can be discovered in time, providing a scientific basis for fault prevention and maintenance, and extending the service life of the display unit; this technical solution can identify and locate various types of faults and has a wide range of applicability; with the continuous advancement of image processing technology and machine learning algorithms, this technical solution can be easily upgraded and optimized to adapt to new fault types and detection requirements.

[0042] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit, characterized in that: The method comprises: S1, collecting images of the narrow-frame TFT-LCD display unit through a high-resolution camera; and preprocessing the collected images; S2. Extract the fault features in the display unit based on image processing technology, analyze the pixels in the image point by point by setting a specific algorithm, identify abnormal pixels or areas, and extract feature information; S3. Based on the machine learning algorithm, a fault classification and location model is constructed, and a large number of display unit images with known faults are used as training samples to train the model; during the training process, the parameters of the model are continuously optimized; S4. Send a specific test signal to the display unit and observe its response to determine whether there is a fault; and use high-precision measuring instruments to accurately measure the fault point; S5. According to the results of image analysis and hardware detection, determine the specific location of the fault point and give corresponding repair suggestions.

2. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 1, characterized in that: Said S1 comprises: S11, using a high-resolution camera to perform all-round image acquisition of the narrow-frame TFT-LCD display unit to capture every detail of the display unit; S12, based on adaptive exposure technology, automatically adjusts the exposure parameters of the camera according to the ambient light, captures images under different lighting conditions, and applies automatic white balance technology to eliminate color deviation; S13, preprocessing the collected image, wherein the preprocessing includes denoising, contrast enhancement, and edge sharpening.

3. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 1, characterized in that: The S2 comprises: S21, refining the fault characteristics of the TFT-LCD display unit; and after image preprocessing, performing preliminary screening through image statistical characteristics to exclude images that are obviously normal or too abnormal; S22, for each pixel, calculating the difference between it and surrounding pixels, and judging whether there is an abnormality based on a preset threshold; and recording the position of the abnormal pixel, the difference value and the statistical characteristics of the surrounding pixels; S23, performing neighborhood analysis on the initially identified abnormal pixels to determine whether they constitute a specific fault mode; further confirming the existence of the fault by calculating the abnormal pixel features in the neighborhood; S24, extracting detailed feature information from the confirmed fault area, and performing comparison and classification using a preset fault feature library based on the extracted feature information; S25, further processing the fault area using morphological operations, and performing context analysis based on multiple factors, and identifying the fault mode that depends on the specific background or environment based on the analysis results; S26. Summarize all the extracted fault features and their classification results to form a fault feature list or a fault distribution map, and verify the extracted fault features using known fault samples or manual inspection.

4. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 3, characterized in that: The S25 comprises: S251, performing morphological processing on the fault area, and performing refinement processing on the fault area based on the morphological processing; S252, analyzing the relative position relationship between the fault area and the edge, center or other key areas of the image based on the position of the fault area in the entire image; S253, analyzing the relative relationship between the fault area and other normal or abnormal areas, and analyzing whether the fault area is located in a specific pattern, icon or text area in combination with the overall layout of the image; S254, extracting features of the fault area using a deep learning network, dividing the image into multiple areas using an image segmentation algorithm, and performing fault detection on each area; S255. For dynamic flicker faults, the time series analysis technology is combined to capture the changing rules of the faults, and the extracted fault features are verified using known fault samples, and the wrong or omitted fault features are corrected and supplemented; S256. Feedback the verification results to the previous processing steps to optimize algorithm parameters, improve processing strategies, and enhance feature extraction capabilities.

5. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 4, characterized in that: The S254 includes: According to the characteristics of the fault area, a deep learning model is selected, and feature extraction is performed based on the selected deep learning model; Collect fault image samples, annotate them, identify the location and type of the fault area, and train the deep learning model to learn the feature representation of the fault area; According to the complexity of the fault area and the overall layout of the image, the image segmentation algorithm is selected and the segmentation results are post-processed; Use the trained deep learning model to extract features from the segmented fault area, and perform feature selection and dimensionality reduction operations; The extracted feature vector is input into the classifier to detect the fault area, and the detected fault area is further classified; According to the feature vector of the fault area and the known fault type library, a classifier is used to determine the specific type of the fault; The detection results are verified using known fault samples and fed back to the deep learning model, image segmentation algorithm, and feature extraction and classification steps.

6. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 4, characterized in that: The S255 includes: Capture the occurrence process of dynamic flicker faults through high-speed camera equipment or video recording technology, and pre-process the captured dynamic fault data; The periodic characteristics of dynamic flicker faults are analyzed by Fourier transform. By identifying the frequency and period of fault occurrence, the correlation between the fault and the external environment or internal factors of the system is obtained. Use trend line fitting to analyze the long-term trend of dynamic flicker faults; combine statistical methods and machine learning algorithms to detect abnormal points or mutation points in dynamic flicker faults; Based on the results of time series analysis, the key features of dynamic flicker faults are extracted, and the extracted fault features are verified using known fault samples; By comparing the matching degree between the actual fault features and the preset feature library, the accuracy and reliability of feature extraction are evaluated, and the features that do not meet expectations are corrected and supplemented; According to the verification results, the feature extraction algorithm is optimized, and the verified fault characteristics are fed back to the time series analysis step and the feature extraction algorithm. According to the feedback results, the time series analysis algorithm and the feature extraction algorithm are iteratively optimized.

7. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 1, characterized in that: The S3 includes: S31, obtaining image data of a narrow-frame TFT-LCD display unit with a known fault, and annotating the collected image data in detail; S32, dividing the labeled data set into a training set, a validation set and a test set, wherein the training set is used for model training, the validation set is used for model tuning and parameter selection, and the test set is used to evaluate the final performance of the model; S33. Select a machine learning model based on the characteristics of the fault detection and classification tasks; S34. Perform preliminary training on the model using the training set. According to the results of the preliminary training, adjust and optimize the model's hyperparameters, and use regularization techniques to prevent the model from overfitting. At the same time, evaluate the generalization ability of the model by monitoring the performance of the validation set, and stop training in advance. S35. During the training process, the input data is continuously enhanced and the performance of the model is evaluated. Finally, the model is evaluated through the test set. S36. Based on the evaluation results of the test set, further optimize and improve the model, retrain the optimized model, and repeat the above evaluation and verification process until the model performance meets the requirements.

8. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 1, characterized in that: The S4 comprises: S41, the test platform automatically starts the image acquisition device to acquire high-resolution images of the narrow-frame TFT-LCD display unit; and automatically transmits the acquired images to the image processing system. S42, the image acquisition system analyzes the image based on a pre-trained machine learning model to identify potential fault areas and types; S43. According to the result of the image analysis, the test platform automatically selects the corresponding electrical test instrument to perform an electrical performance test on the display unit; S44. Combining the results of the image analysis and the electrical performance test, the test platform automatically locates the fault area and type in the display unit; and automatically generates a fault report.

9. A method for detecting and locating faults in a narrow-frame TFT-LCD display unit according to claim 1, characterized in that: The S5 comprises: S51, collecting detailed fault phenomenon descriptions from the aforementioned detection steps, conducting in-depth analysis on the collected fault data, and identifying the cause of the fault; S52. Classify the fault into different types according to the fault phenomenon and data, where the different types include hardware fault, software fault and design defect; S53. Based on the fault classification and fault phenomenon, preliminarily determine which subsystem or component the fault may occur in, and conduct a detailed investigation of the preliminarily located subsystem or component using a test instrument; S54, gradually narrowing down the scope of the fault by step-by-step elimination until the specific fault point is located. After the fault point is located, verifying whether the fault exists by simulating the fault condition or replacing the faulty component; S55. In combination with the results of fault location and the design principle of the display unit, the cause of the fault is analyzed in depth, and the accuracy and reliability of the cause of the fault are confirmed through further testing and verification; S56. Develop a specific repair plan based on the cause of the failure and conduct a cost assessment on the repair plan; S57. If the repair cost is too high or the repair difficulty is great, consider whether there are alternative solutions, such as using a spare display unit or upgrading the display unit, and repair the display unit according to the repair plan; S58. After the repair is completed, the display unit is subjected to a function test and a performance test, and the repair process, repair plan and repair result are recorded.

10. A narrow-frame TFT-LCD display unit fault detection and positioning system, characterized in that: The system comprises: Data acquisition module: collects images of the narrow-frame TFT-LCD display unit through a high-resolution camera and pre-processes the collected images; Feature extraction module: Based on image processing technology, it extracts the fault features in the display unit. By setting a specific algorithm, it analyzes the pixels in the image point by point, identifies abnormal pixels or areas, and extracts feature information. Model training module: Based on the machine learning algorithm, a fault classification and location model is constructed, and a large number of display unit images with known faults are used as training samples to train the model; during the training process, the model parameters are continuously optimized; Fault judgment module: by sending a specific test signal to the display unit and observing its response, it can be judged whether there is a fault; and the fault point can be accurately measured using high-precision measuring instruments; Suggestion feedback module: Determine the specific location of the fault point based on the results of image analysis and hardware detection, and give corresponding repair suggestions.

Images