OLED display area dark line defect identification method and system based on image processing

Abstract

The invention discloses an OLED display area dark line defect identification method and system based on image processing, and relates to the technical field of image processing, and the method comprises the steps: calculating the amplitude-frequency response abnormality of an amplitude-frequency response diagram, calculating the group delay abnormality of a group delay diagram, fusing the amplitude-frequency response abnormality and the group delay abnormality, and obtaining the dark line defect of an OLED display area. Generating a final channel response anomaly feature; and creating a binarized image with the same size as the ridge line intensity graph, constructing a covariance matrix and carrying out eigendecomposition, defining a unit vector of an eigenvector corresponding to the maximum eigenvalue as a dominant direction of a defect segment, and superposing and drawing the unit vector on the phase-shift sine stripe test image to generate a defect detection result graph. According to the method, through frequency domain transformation and group delay evaluation, the sensitivity to fine defects is improved, and the identification capability of dark line defects in a complex background is enhanced by adopting fusion of amplitude-frequency response anomaly and group delay anomaly.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and system for identifying dark line defects in OLED display areas based on image processing. Background Technology

[0002] With the rapid development of display technology, OLED (Organic Light Emitting Diode) display technology has gradually become one of the mainstream display technologies due to its excellent display effect and broad application prospects. The advantages of OLED displays are their ultra-thinness, flexibility, low power consumption and high contrast, and they are widely used in many fields such as televisions, smartphones, and wearable devices. OLED displays are prone to some defects during production and use, especially dark line defects in the display area. These defects can significantly affect display quality and user experience, so effective detection and identification are particularly important.

[0003] Despite the shortcomings of existing technologies in OLED display defect detection, traditional vision-based detection methods are subject to noise interference and have difficulty accurately identifying small-area dark line defects in complex backgrounds. Existing technologies still face the problem of not being able to fully exploit frequency domain features and group delay information, resulting in insufficient defect recognition accuracy in complex backgrounds. In particular, the defect recognition effect is often unsatisfactory under high noise or low contrast conditions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for identifying dark line defects in OLED display areas based on image processing. It solves the problems of traditional vision-based detection methods being affected by noise interference and having difficulty accurately identifying small-area dark line defects in complex backgrounds. Existing technologies still face the problem of not being able to fully exploit frequency domain features and group delay information, resulting in insufficient defect identification accuracy in complex backgrounds. In particular, the defect identification effect is often unsatisfactory under high noise or low contrast conditions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for identifying dark line defects in OLED display areas based on image processing, comprising,

[0008] Acquire phase-shifted sinusoidal fringe images from an OLED display and preprocess them. Extract grayscale values ​​from the phase-shifted sinusoidal fringe test images to generate time-domain signals. Perform frequency domain transformation on the time-domain signals to obtain complex spectra. Define the magnitude of the complex spectrum at the index as the amplitude-frequency response. Calculate the index corresponding to the frequency and evaluate the group delay at the index.

[0009] The amplitude frequency response and group delay are arranged according to the pixel coordinates of the phase-shifted sinusoidal fringe test image to form the amplitude frequency response map and the group delay map respectively. The amplitude frequency response anomaly of the amplitude frequency response map is calculated, and the group delay anomaly of the group delay map is calculated. The amplitude frequency response anomaly and the group delay anomaly are fused to generate the final channel response anomaly feature.

[0010] Calculate the first-order partial derivatives of the channel response anomaly characteristics and calculate the outer product, which is defined as the gradient structure tensor. Perform two-dimensional Gaussian filtering and eigenvalue decomposition on the gradient structure tensor to obtain eigenvalues ​​and corresponding eigenvectors. Define the ridge strength based on the coherence measure of the eigenvalues ​​and the ridge strength based on the eigenvectors.

[0011] Create a binarized image with the same size as the ridge intensity map, generate connected components, calculate the center of all pixels in the connected components, construct the covariance matrix and perform eigenvalue decomposition, define the unit vector of the eigenvector corresponding to the largest eigenvalue as the dominant direction of the defect segment, and overlay it on the phase-shifted sinusoidal fringe test image to generate a defect detection result map.

[0012] As a preferred embodiment of the image processing-based OLED display dark line defect identification method of the present invention, the steps of extracting gray values ​​from the phase-shifted sinusoidal stripe test image, generating a time-domain signal, performing frequency domain transformation on the time-domain signal to obtain a complex spectrum, defining the magnitude of the complex spectrum at the index as the amplitude-frequency response, calculating the index corresponding to the frequency, and evaluating the group delay at the index include:

[0013] For each pixel coordinate on the imaging plane of the OLED display, the gray value at the corresponding position in the phase-shifted sinusoidal fringe test image is extracted to generate a time-domain signal;

[0014] Perform a frequency domain transformation on the time domain signal of each pixel to obtain a complex spectrum, and define the magnitude of the complex spectrum at the index as the amplitude-frequency response;

[0015] In the complex spectrum, the index corresponding to the frequency is calculated, and the group delay at the index is evaluated using the central difference method.

[0016] As a preferred embodiment of the image processing-based OLED display region dark line defect identification method of the present invention, wherein: the fusion of amplitude-frequency response anomaly and group delay anomaly to generate the final channel response anomaly feature includes:

[0017] The amplitude-frequency response and group delay are arranged according to the pixel coordinates of the phase-shifted sinusoidal fringe test image to form the amplitude-frequency response map and the group delay map, respectively.

[0018] Morphological opening is performed on the amplitude-frequency response map to obtain the background estimation map. The degree of attenuation of each pixel relative to the background estimation map is calculated and defined as the amplitude-frequency response anomaly.

[0019] Based on the group delay map, the group delay anomaly of each pixel is calculated using the Z-score normalization method;

[0020] The amplitude-frequency response anomaly and the group delay anomaly are fused using geometric averaging to generate the final channel response anomaly feature.

[0021] As a preferred embodiment of the image processing-based OLED display region dark line defect identification method of the present invention, the following steps are included: calculating the first-order partial derivative of the channel response anomaly characteristic and calculating the outer product, defined as a gradient structure tensor; performing two-dimensional Gaussian filtering and eigenvalue decomposition on the gradient structure tensor to obtain eigenvalues ​​and corresponding eigenvectors; defining ridge line intensity based on the coherence measure of the eigenvalues; and defining ridge line intensity based on the eigenvectors, including:

[0022] Calculate the first-order partial derivatives of the channel response anomaly characteristics to obtain the gradient components in the horizontal and vertical directions. Calculate the outer product of the gradient components to obtain a symmetric positive semi-definite matrix, which is defined as the gradient structure tensor.

[0023] Two-dimensional Gaussian filtering is applied to each component of the gradient structure tensor to obtain a smoothed tensor field. Eigenvalues ​​and corresponding eigenvectors are then obtained from the smoothed gradient structure tensor field.

[0024] Ridge strength is defined based on a coherence measure of eigenvalues.

[0025] The eigenvector corresponding to the largest eigenvalue is denoted as the principal eigenvector. The principal eigenvector is then rotated by 90 degrees. The direction behind is defined as the ridgeline direction;

[0026] Set a low threshold and filter each pixel whose ridge intensity is greater than the low threshold. Calculate the gradient direction of the ridge. Along the gradient direction, at a distance of 1 pixel, calculate the coordinates of two interpolation points. Use bilinear interpolation to obtain the ridge intensity at the two interpolation points from the ridge intensity. If the ridge intensity of the current pixel is greater than or equal to the ridge intensity at the two interpolation points, retain the ridge intensity of the current pixel; otherwise, set it to zero to generate the thinned ridge intensity.

[0027] As a preferred embodiment of the image processing-based OLED display area dark line defect identification method of the present invention, wherein: the construction of the covariance matrix and the eigenvalue decomposition, defining the unit vector of the eigenvector corresponding to the largest eigenvalue as the dominant direction of the defect segment, includes:

[0028] Create a binary image of the same size as the thinned ridge intensity map, with all initial values ​​set to 0. In the thinned ridge intensity map, find the unfiltered pixel with the largest ridge intensity value and use it as the current seed point. Starting from the current seed point, perform directional constraint region growing. If the constraint conditions are met, merge the neighboring pixels into the current region and mark them as visited. Continue until all pixels with ridge intensity greater than 0 have been marked, thus obtaining the connected component.

[0029] The region growth with directional constraints includes initializing an empty region point set, adding the current seed point to the region point set and marking it, and checking the 8-neighbor pixels of all pixels in the point set.

[0030] The constraints include that the ridge intensity of neighboring pixels is greater than a low threshold, and the absolute difference between the ridge direction of neighboring pixels and the ridge direction of the seed pixel is less than a direction threshold.

[0031] Calculate the center of all pixels in the connected domain, construct the centered coordinate matrix, calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and define the unit vector of the eigenvector corresponding to the largest eigenvalue as the dominant direction of the defect segment.

[0032] Project all pixels within the connected domain onto the dominant direction, calculate the projection scalar, find the minimum and maximum values ​​among all projection scalars, and inversely transform the corresponding projection points back to the phase-shifted sinusoidal fringe test image coordinate system to obtain the two endpoints of the line segment.

[0033] As a preferred embodiment of the image processing-based OLED display area dark line defect identification method of the present invention, wherein: the step of superimposing and drawing on the phase-shifted sinusoidal fringe test image to generate a defect detection result image includes:

[0034] The dominant direction and endpoints of the defect fragments are superimposed on the phase-shifted sinusoidal fringe test image to generate a defect detection result image.

[0035] As a preferred embodiment of the image processing-based OLED display area dark line defect identification method of the present invention, the step of acquiring a phase-shifted sinusoidal stripe image of the OLED display screen and performing preprocessing includes:

[0036] Phase-shifted sinusoidal fringe test images were acquired using an industrial camera and then denoised and normalized.

[0037] Secondly, the present invention provides an OLED display area dark line defect identification system based on image processing, comprising,

[0038] The acquisition and processing module is used to acquire phase-shifted sinusoidal fringe images and perform noise reduction and normalization processing;

[0039] The frequency domain analysis module is used to extract the gray value of each pixel, generate a time domain signal, and perform a frequency domain transformation to obtain a complex spectrum, thereby calculating the amplitude frequency response and group delay.

[0040] The anomaly graph module is used to calculate the amplitude-frequency response anomaly through morphological operations and the group delay anomaly through Z-score normalization, and obtain the anomaly graphs of the two.

[0041] The fusion direction module is used to fuse the amplitude-frequency response anomaly and the group delay anomaly through geometric mean to obtain the final channel response anomaly feature, calculate the gradient structure tensor, obtain the ridge intensity map through feature decomposition, and perform directionally constrained region growth according to the ridge direction.

[0042] The defect detection module is used to locate defect segments using the region growing method, calculate the projection scalar, obtain the endpoint coordinates of the defect line segments, and superimpose the dominant direction of the defect segment and the fitted line segment endpoints onto the original image to generate a defect detection result image.

[0043] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the image processing-based OLED display area dark line defect identification method as described in the first aspect of the present invention.

[0044] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the image processing-based OLED display area dark line defect identification method as described in the first aspect of the present invention.

[0045] The beneficial effects of this invention are as follows: By using frequency domain transformation and group delay evaluation, this invention can extract valuable spectral and temporal features from deeper image information, thereby reducing noise interference and improving sensitivity to subtle defects. By using the fusion of amplitude-frequency response anomaly and group delay anomaly, it can effectively enhance the ability to identify dark line defects in complex backgrounds. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1This is a flowchart of the OLED display area dark line defect identification method based on image processing in Example 1.

[0048] Figure 2 This is a schematic diagram of the OLED display area dark line defect identification system based on image processing in Example 1. Detailed Implementation

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0052] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a method for identifying dark line defects in OLED display areas based on image processing, including the following steps:

[0053] S1. Acquire the phase-shifted sinusoidal fringe image of the OLED display and perform preprocessing. Extract the gray values ​​from the phase-shifted sinusoidal fringe test image to generate a time-domain signal. Perform frequency domain transformation on the time-domain signal to obtain a complex spectrum. Define the magnitude of the complex spectrum at the index as the amplitude-frequency response. Calculate the index corresponding to the frequency and evaluate the group delay at the index.

[0054] Specifically, the phase-shifted sinusoidal fringe image of the OLED display is acquired and preprocessed, including:

[0055] The OLED display under test is placed in an optical darkroom. An industrial camera is fixed directly in front of the screen using an adjustment bracket. The camera's orientation is adjusted so that its optical axis is perpendicular to the screen plane, ensuring the camera's field of view covers the entire screen area. The industrial camera is then used to acquire a phase-shifted sinusoidal fringe image displayed on the OLED display, which is then denoised and normalized using the following formula:

[0056] ,

[0057] in Let represent the k-th phase-shifted sinusoidal fringe test image, and let represent the grayscale value of the k-th frame of the phase-shifted sinusoidal fringe test image at coordinates (u, v), where u and v are the column index and row index of the pixel, respectively. The average gray level is set using the median of an 8-bit image, and m is the modulation depth, which is set using an experimental calibration method (by taking test images with different m values, the m value with the highest signal-to-noise ratio and no saturation is selected). Let K be the fringe spatial frequency, and K be the phase shift step number. K=4 is set using the four-step phase shift method.

[0058] The preprocessing step effectively improves image quality and reduces the impact of noise. By precisely controlling the acquisition conditions, it ensures the comparability and consistency between different test images, providing more accurate data for defect identification.

[0059] Furthermore, grayscale values ​​are extracted from the phase-shifted sinusoidal fringe test image to generate a time-domain signal. A frequency-domain transformation is performed on the time-domain signal to obtain a complex spectrum. The magnitude of the complex spectrum at the index is defined as the amplitude-frequency response. The index corresponding to the frequency is calculated, and the group delay at the index is evaluated, including:

[0060] For each pixel coordinate on the imaging plane of the OLED display, the gray value at the corresponding position in the phase-shifted sinusoidal fringe test image is extracted to generate a time-domain signal;

[0061] For each pixel's time-domain signal, a frequency-domain transformation is performed (first, a Hanning window function is applied to the time-domain signal to obtain a windowed signal, and a discrete Fourier transform is performed on the windowed signal to obtain its complex spectrum), resulting in a complex spectrum. The magnitude of the complex spectrum at the index is defined as the amplitude-frequency response.

[0062] In a complex spectrum, the index corresponding to a frequency is calculated using the following formula:

[0063] ,

[0064] in Excitation frequency The integer index corresponding to the complex number spectrum The time-domain frequency of the phase-shift fringe excitation signal is... The signal sampling rate is the frame rate at which the industrial camera captures images. A function that rounds the integer to the nearest whole number;

[0065] The central difference method is used to evaluate the group latency at the index, with the following formula:

[0066] ,

[0067] ,

[0068] ,

[0069] in For pixel (u,v) at the excitation frequency Group delay estimate at [location] The phase spectrum is a complex spectrum. For the frequency resolution of the complex spectrum, The complex spectrum represents the pixel spectrum at pixel (u,v), specifically the complex value at the frequency index n.

[0070] Frequency domain transformation of time-domain signals enables the extraction of frequency features from images, and frequency domain analysis helps identify periodic noise and defect signals in images. Combined analysis of group delay and amplitude-frequency response allows for efficient identification of signal changes hidden in complex backgrounds within the frequency domain, especially for defects in low-contrast and high-noise conditions. By combining the anomalies of amplitude-frequency response and group delay, the accuracy of detection and sensitivity to minute defects are greatly improved. The calculation of gradient structure tensor and ridge intensity analysis make the fine structures in the image clearer, especially in the location and identification of minute defects in display panels, which can significantly improve accuracy. The location and projection analysis of defect segments ensure the accuracy and reliability of defect detection, accurately identifying dark line defects in the display area and accurately locating the defect position.

[0071] S2. Arrange the amplitude frequency response and group delay according to the pixel coordinates of the phase-shifted sinusoidal fringe test image to form the amplitude frequency response map and the group delay map respectively. Calculate the amplitude frequency response anomaly of the amplitude frequency response map and the group delay anomaly of the group delay map. Fuse the amplitude frequency response anomaly and the group delay anomaly to generate the final channel response anomaly feature.

[0072] Specifically, the amplitude-frequency response anomaly and group delay anomaly are fused to generate the final channel response anomaly feature, including:

[0073] The amplitude-frequency response and group delay are arranged according to the pixel coordinates of the phase-shifted sinusoidal fringe test image to form the amplitude-frequency response map and the group delay map, respectively.

[0074] A morphological opening operation (erosion followed by dilation) is performed on the amplitude-frequency response map to obtain a background estimation map. The degree of attenuation of each pixel relative to the background estimation map is calculated and defined as the amplitude-frequency response anomaly, as shown in the formula:

[0075] ,

[0076] in This is a diagram showing the anomaly of the amplitude-frequency response. This is a background estimation map of the amplitude-frequency response obtained through morphological opening operations. This is the amplitude-frequency response diagram. It is a very small positive number used to prevent the denominator from being zero and to ensure numerical stability;

[0077] Based on the group delay map, the group delay anomaly of each pixel is calculated using the Z-score normalization method, with the following formula:

[0078] ,

[0079] in For group delay anomaly map, For group delay graphs, and The mean and standard deviation of the group delay;

[0080] The amplitude-frequency response anomaly and group delay anomaly are fused using a geometric mean to generate the final channel response anomaly feature, as shown in the formula:

[0081] ,

[0082] in The final channel response anomaly characteristic, This is a group delay anomaly graph.

[0083] By geometrically fusing the amplitude-frequency response anomaly and the group delay anomaly, we can more accurately capture the characteristics of defects, especially in environments with complex backgrounds and noise. The fused channel response anomaly characteristics not only enhance the sensitivity to minor defects but also improve the robustness of the system, effectively reducing the impact of noise interference on the results.

[0084] S3. Calculate the first-order partial derivatives of the channel response anomaly characteristics and calculate the outer product, which is defined as the gradient structure tensor. Perform two-dimensional Gaussian filtering and eigenvalue decomposition on the gradient structure tensor to obtain eigenvalues ​​and corresponding eigenvectors. Define the ridge strength based on the coherence measure of the eigenvalues ​​and the ridge strength based on the eigenvectors.

[0085] Specifically, the first-order partial derivatives of the channel response anomaly characteristics are calculated, and the outer product is calculated, defined as the gradient structure tensor. Two-dimensional Gaussian filtering and eigenvalue decomposition are performed on the gradient structure tensor to obtain eigenvalues ​​and corresponding eigenvectors. Based on the coherence measure of the eigenvalues, the ridge strength is defined. Based on the eigenvectors, the ridge strength is further defined, including:

[0086] The first-order partial derivatives of the channel response anomaly characteristics are approximated using the central difference method, yielding gradient components along the horizontal (u-axis) and vertical (v-axis) directions. The outer product of these gradient components is then calculated to obtain a symmetric positive semi-definite matrix, defined as the gradient structure tensor, with the following formula:

[0087] ,

[0088] in For gradient structure tensors, and These are the gradient components along the horizontal (u-axis) and vertical (v-axis) directions;

[0089] Two-dimensional Gaussian filtering is applied to each component of the gradient structure tensor to fuse gradient information within the local neighborhood, improving robustness to noise and obtaining a more coherent structure estimate, resulting in a smoothed tensor field, as shown in the formula:

[0090] ,

[0091] in This is the gradient structure tensor field after Gaussian smoothing. The standard deviation is Two-dimensional Gaussian convolution kernel, To determine the standard deviation of the Gaussian kernel and control the smoothing scale, a scale selection method is used. This is a two-dimensional discrete convolution operation;

[0092] Eigenvalues ​​and corresponding eigenvectors are obtained by performing eigendecomposition on the smoothed gradient structure tensor field.

[0093] Based on the coherence measure of eigenvalues, ridge strength is defined by the following formula:

[0094] ,

[0095] in This is a ridge intensity map. and These are the eigenvalues ​​of the gradient structure tensor, reflecting the magnitudes of the principal and secondary components of the local gradient energy;

[0096] The eigenvector corresponding to the largest eigenvalue is denoted as the principal eigenvector. The principal eigenvector is then rotated by 90 degrees. The direction after is defined as the ridge direction, and the result is normalized to an interval. To eliminate periodic ambiguity in direction and angle, the formula is:

[0097] ,

[0098] in The ridge direction field represents the direction angle of the ridge tangent at that point. and These are the projection components of the principal feature vector onto the x-axis (horizontal direction) and y-axis (vertical direction);

[0099] Using the percentile method, a low threshold is set at 75% of the ridge intensity. Pixels with ridge intensity greater than this low threshold are then selected. The gradient direction (ridge direction) is then calculated. The calculation is performed using the orthogonal direction transformation method. (Based on the ridge direction of the point, its normal direction, i.e., the gradient direction, is determined.) Along the gradient direction (one direction in the forward direction and one in the reverse direction), at a distance of 1 pixel, the coordinates of two interpolation points are calculated. Using the bilinear interpolation method, the ridge intensity at the two interpolation points is obtained from the ridge intensity. If the ridge intensity of the current pixel is greater than or equal to the ridge intensity at the two interpolation points, the ridge intensity of the current pixel is retained; otherwise, it is set to zero, generating the thinned ridge intensity.

[0100] By analyzing local gradient information, defect detection can not only reflect the overall frequency domain characteristics but also provide more precise local features, thereby improving the accuracy of defect localization. Through the calculation of ridge intensity and the determination of ridge direction, it is possible to effectively describe the fine defect structure on the display panel, especially for dark line defects, providing highly identifiable directional information to ensure accurate localization in the micro-defect region. Furthermore, by determining the projection of the minimum and maximum values, the accurate identification of defect segments is further enhanced. The precise calculation of the dominant direction enables defect detection to not only identify the location of the defect but also provide specific directional information, resulting in more accurate localization of the defect region. It can also effectively identify and analyze micro-linear defects on the display panel. The refinement and localization process of channel response anomaly characteristics can significantly improve the accuracy of defect segment localization, especially when dealing with more complex display panel defects, providing accurate defect boundaries.

[0101] S4. Create a binarized image with the same size as the ridge intensity map, generate connected components, calculate the center of all pixels in the connected components, construct the covariance matrix and perform eigenvalue decomposition, define the unit vector of the eigenvector corresponding to the largest eigenvalue as the dominant direction of the defect segment, and overlay it on the phase-shifted sinusoidal fringe test image to generate a defect detection result map.

[0102] Specifically, the covariance matrix is ​​constructed and eigenvalue decomposition is performed. The unit vector of the eigenvector corresponding to the largest eigenvalue is defined as the dominant direction of the defect segment, including:

[0103] Create a binary image with the same size as the thinned ridge intensity map (binarization image = 0 indicates that the pixel is not marked, and binarization image = 1 indicates that it has been marked). The initial values ​​are all 0. In the thinned ridge intensity map, find the pixel that has not been filtered and has the largest ridge intensity value. Use it as the current seed point. Starting from the current seed point, perform directional constraint region growing. If the constraint conditions are met, the neighboring pixels are merged into the current region and marked as visited. Continue until all pixels with ridge intensity greater than 0 have been marked, and a connected component is obtained.

[0104] The region growth with directional constraints includes initializing an empty region point set, adding the current seed point to the region point set and marking it, and checking the 8-neighbor pixels of all pixels in the point set.

[0105] The constraints include the ridge intensity of neighboring pixels being greater than a low threshold, and the absolute difference between the ridge direction of neighboring pixels and the ridge direction of the seed point (normalized to...). If the value is less than the direction threshold, the direction threshold should be set using an empirical threshold method.

[0106] Calculate the center of all pixels in the connected domain, construct the centered coordinate matrix, calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and define the unit vector of the eigenvector corresponding to the largest eigenvalue as the dominant direction of the defect segment.

[0107] Projecting all pixels within the connected component onto the dominant direction, the projection scalar is calculated using the following formula:

[0108] ,

[0109] in For the j-th projected scalar, Let J be the coordinates of the j-th pixel within the i-th connected component. Let i be the center of all pixels in the i-th connected component. The dominant direction is the unit vector;

[0110] Find the minimum and maximum values ​​among all projected scalars, and inversely transform the corresponding projected points back to the phase-shifted sinusoidal fringe test image coordinate system to obtain the two endpoints of the line segment. The formula is:

[0111] ,

[0112] ,

[0113] in and Let be the coordinates of the start and end points of the line segment. and These are the minimum and maximum values ​​in the projected scalar.

[0114] Binarization and connected component generation can effectively extract continuous defect regions from complex images and effectively remove noise during the process, ensuring that subsequent analysis is based on valid regions. This directional constraint can effectively guarantee the directionality and consistency of region growth, avoiding over-expansion or misidentification of defect regions caused by erroneous growth in incoherent directions. Through eigenvalue decomposition of the covariance matrix, the morphology and structure of defect regions can be accurately analyzed from a geometric perspective, avoiding the limitations of traditional methods that can only rely on brightness or color changes for feature extraction. The calculation of the projection scalar provides a more accurate geometric description of defect fragments, especially in the detection of small or blurry defects, ensuring the clarity and accuracy of defect boundaries.

[0115] Furthermore, the image is overlaid onto the phase-shifted sinusoidal fringe test image to generate a defect detection result map, including:

[0116] The dominant direction of the defect fragment and the endpoints of the fitted line segments are superimposed on the phase-shifted sinusoidal fringe test image to generate a defect detection result image.

[0117] Through the final defect detection result image, users can intuitively see the location and shape of the defect, which provides necessary support for further analysis, processing and repair work.

[0118] This embodiment also provides an OLED display area dark line defect identification system based on image processing, including:

[0119] The acquisition and processing module is used to acquire phase-shifted sinusoidal fringe images and perform noise reduction and normalization processing;

[0120] The frequency domain analysis module is used to extract the gray value of each pixel, generate a time domain signal, and perform a frequency domain transformation to obtain a complex spectrum, thereby calculating the amplitude frequency response and group delay.

[0121] The anomaly graph module is used to calculate the amplitude-frequency response anomaly through morphological operations and the group delay anomaly through Z-score normalization, and obtain the anomaly graphs of the two.

[0122] The fusion direction module is used to fuse the amplitude-frequency response anomaly and the group delay anomaly through geometric mean to obtain the final channel response anomaly feature, calculate the gradient structure tensor, obtain the ridge intensity map through feature decomposition, and perform directionally constrained region growth according to the ridge direction.

[0123] The defect detection module is used to locate defect segments using the region growing method, calculate the projection scalar, obtain the endpoint coordinates of the defect line segments, and superimpose the dominant direction of the defect segment and the fitted line segment endpoints onto the original image to generate a defect detection result image.

[0124] This embodiment also provides a computer device applicable to the image processing-based method for identifying dark line defects in OLED display areas, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the image processing-based method for identifying dark line defects in OLED display areas as proposed in the above embodiment.

[0125] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0126] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the image processing-based method for identifying dark line defects in OLED display areas as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0127] In summary, this invention, through frequency domain transformation and group delay evaluation, can extract valuable spectral and temporal features from deeper image information, thereby reducing noise interference and improving sensitivity to subtle defects. By fusing amplitude-frequency response anomaly and group delay anomaly, it can effectively enhance the ability to identify dark line defects in complex backgrounds.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying dark line defects in OLED display areas based on image processing, characterized in that: include, Acquire phase-shifted sinusoidal fringe images from an OLED display and preprocess them. Extract grayscale values ​​from the phase-shifted sinusoidal fringe test images to generate time-domain signals. Perform frequency domain transformation on the time-domain signals to obtain complex spectra. Define the magnitude of the complex spectrum at the index as the amplitude-frequency response. Calculate the index corresponding to the frequency and evaluate the group delay at the index. The amplitude frequency response and group delay are arranged according to the pixel coordinates of the phase-shifted sinusoidal fringe test image to form the amplitude frequency response map and the group delay map respectively. The amplitude frequency response anomaly of the amplitude frequency response map is calculated, and the group delay anomaly of the group delay map is calculated. The amplitude frequency response anomaly and the group delay anomaly are fused to generate the final channel response anomaly feature. Calculate the first-order partial derivatives of the channel response anomaly characteristics and calculate the outer product, which is defined as the gradient structure tensor. Perform two-dimensional Gaussian filtering and eigenvalue decomposition on the gradient structure tensor to obtain eigenvalues ​​and corresponding eigenvectors. Define the ridge strength based on the coherence measure of the eigenvalues ​​and the ridge strength based on the eigenvectors. Create a binarized image with the same size as the ridge intensity map, generate connected components, calculate the center of all pixels in the connected components, construct the covariance matrix and perform eigenvalue decomposition, define the unit vector of the eigenvector corresponding to the largest eigenvalue as the dominant direction of the defect segment, and overlay it on the phase-shifted sinusoidal fringe test image to generate a defect detection result map.

2. The method for identifying dark line defects in OLED display areas based on image processing as described in claim 1, characterized in that: The process involves extracting grayscale values ​​from the phase-shifted sinusoidal fringe test image, generating a time-domain signal, performing a frequency-domain transformation on the time-domain signal to obtain a complex spectrum, defining the magnitude of the complex spectrum at an index as the amplitude-frequency response, calculating the index corresponding to the frequency, and evaluating the group delay at the index, including: For each pixel coordinate on the imaging plane of the OLED display, the gray value at the corresponding position in the phase-shifted sinusoidal fringe test image is extracted to generate a time-domain signal; Perform a frequency domain transformation on the time domain signal of each pixel to obtain a complex spectrum, and define the magnitude of the complex spectrum at the index as the amplitude-frequency response; In the complex spectrum, the index corresponding to the frequency is calculated, and the group delay at the index is evaluated using the central difference method.

3. The method for identifying dark line defects in OLED display areas based on image processing as described in claim 2, characterized in that: The process of fusing the amplitude-frequency response anomaly and the group delay anomaly to generate the final channel response anomaly feature includes: The amplitude-frequency response and group delay are arranged according to the pixel coordinates of the phase-shifted sinusoidal fringe test image to form the amplitude-frequency response map and the group delay map, respectively. Morphological opening is performed on the amplitude-frequency response map to obtain the background estimation map. The degree of attenuation of each pixel relative to the background estimation map is calculated and defined as the amplitude-frequency response anomaly. Based on the group delay map, the group delay anomaly of each pixel is calculated using the Z-score normalization method; The amplitude-frequency response anomaly and the group delay anomaly are fused using geometric averaging to generate the final channel response anomaly feature.

4. The method for identifying dark line defects in OLED display areas based on image processing as described in claim 3, characterized in that: The first-order partial derivatives of the channel response anomaly characteristics are calculated, and the outer product is calculated, defined as the gradient structure tensor. Two-dimensional Gaussian filtering and eigenvalue decomposition are performed on the gradient structure tensor to obtain eigenvalues ​​and corresponding eigenvectors. Based on the coherence measure of the eigenvalues, the ridge strength is defined. Based on the eigenvectors, the ridge strength is further defined, including: Calculate the first-order partial derivatives of the channel response anomaly characteristics to obtain the gradient components in the horizontal and vertical directions. Calculate the outer product of the gradient components to obtain a symmetric positive semi-definite matrix, which is defined as the gradient structure tensor. Two-dimensional Gaussian filtering is applied to each component of the gradient structure tensor to obtain a smoothed tensor field. Eigenvalues ​​and corresponding eigenvectors are then obtained from the smoothed gradient structure tensor field. Ridge strength is defined based on a coherence measure of eigenvalues. The eigenvector corresponding to the largest eigenvalue is denoted as the principal eigenvector. The principal eigenvector is then rotated. The direction behind is defined as the ridgeline direction; Set a low threshold and filter each pixel whose ridge intensity is greater than the low threshold. Calculate the gradient direction of the ridge. Along the gradient direction, at a distance of 1 pixel, calculate the coordinates of two interpolation points. Use bilinear interpolation to obtain the ridge intensity at the two interpolation points from the ridge intensity. If the ridge intensity of the current pixel is greater than or equal to the ridge intensity at the two interpolation points, retain the ridge intensity of the current pixel; otherwise, set it to zero to generate the thinned ridge intensity.

5. The method for identifying dark line defects in OLED display areas based on image processing as described in claim 4, characterized in that: The construction of the covariance matrix and the eigenvalue decomposition, wherein the unit vector of the eigenvector corresponding to the largest eigenvalue is defined as the dominant direction of the defect segment, includes: Create a binary image of the same size as the thinned ridge intensity map, with all initial values ​​set to 0. In the thinned ridge intensity map, find the unfiltered pixel with the largest ridge intensity value and use it as the current seed point. Starting from the current seed point, perform directional constraint region growing. If the constraint conditions are met, merge the neighboring pixels into the current region and mark them as visited. Continue until all pixels with ridge intensity greater than 0 have been marked, thus obtaining the connected component. The region growth with directional constraints includes initializing an empty region point set, adding the current seed point to the region point set and marking it, and checking the 8-neighbor pixels of all pixels in the point set. The constraints include that the ridge intensity of neighboring pixels is greater than a low threshold, and the absolute difference between the ridge direction of neighboring pixels and the ridge direction of the seed pixel is less than a direction threshold. Calculate the center of all pixels in the connected domain, construct the centered coordinate matrix, calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and define the unit vector of the eigenvector corresponding to the largest eigenvalue as the dominant direction of the defect segment. Project all pixels within the connected domain onto the dominant direction, calculate the projection scalar, find the minimum and maximum values ​​among all projection scalars, and inversely transform the corresponding projection points back to the phase-shifted sinusoidal fringe test image coordinate system to obtain the two endpoints of the line segment.

6. The method for identifying dark line defects in OLED display areas based on image processing as described in claim 5, characterized in that: The superimposed image drawn on the phase-shifted sinusoidal fringe test image generates a defect detection result map, including: The dominant direction and endpoints of the defect fragments are superimposed on the phase-shifted sinusoidal fringe test image to generate a defect detection result image.

7. The method for identifying dark line defects in OLED display areas based on image processing as described in claim 1, characterized in that: The acquisition and preprocessing of the phase-shifted sinusoidal fringe image of the OLED display screen includes: Phase-shifted sinusoidal fringe test images were acquired using an industrial camera and then denoised and normalized.

8. A system for identifying dark line defects in OLED display areas based on image processing, wherein the method for identifying dark line defects in OLED display areas based on image processing is described in any one of claims 1 to 7, characterized in that: include, The acquisition and processing module is used to acquire phase-shifted sinusoidal fringe images and perform noise reduction and normalization processing; The frequency domain analysis module is used to extract the gray value of each pixel, generate a time domain signal, and perform a frequency domain transformation to obtain a complex spectrum, thereby calculating the amplitude frequency response and group delay. The anomaly graph module is used to calculate the amplitude-frequency response anomaly through morphological operations and the group delay anomaly through Z-score normalization, and obtain the anomaly graphs of the two. The fusion direction module is used to fuse the amplitude-frequency response anomaly and the group delay anomaly through geometric mean to obtain the final channel response anomaly feature, calculate the gradient structure tensor, obtain the ridge intensity map through feature decomposition, and perform directionally constrained region growth according to the ridge direction. The defect detection module is used to locate defect segments using the region growing method, calculate the projection scalar, obtain the endpoint coordinates of the defect line segments, and superimpose the dominant direction of the defect segment and the fitted line segment endpoints onto the original image to generate a defect detection result image.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the image processing-based OLED display area dark line defect identification method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the image processing-based OLED display area dark line defect identification method according to any one of claims 1 to 7.

Images