A method and device for detecting sub-pixel defects with adaptive multi-depth of field

Abstract

The present invention discloses an adaptive multi-depth sub-pixel layer defect detection method, comprising the following steps: acquiring pictures of the sub-pixel layer through an image acquisition module to obtain images of different depths of field; preprocessing the acquired images based on an image preprocessing module, specifically adopting a plurality of filters to perform noise reduction and smoothing on the input images to eliminate noise interference and improve image quality; accurately extracting edge features in images of different depths of field through a defect detection module; calculating multi-dimensional feature information of the images through a feature calculation module; designing defect evaluation indicators of human visual perception based on the multi-dimensional feature information obtained above, accurately locating the clearest layer where the defects are located by adaptively adjusting the weights of each evaluation indicator and combining the depth information of the multi-depth images, thereby determining the optimal distribution position of the defects in the sub-pixel layer, and introducing a comparative analysis mechanism consistent with human visual perception to ensure that the detection results are highly consistent with human visual perception.

Description

Technical Field

[0001] The present invention relates to the field of high-precision defect detection in the electronic screen manufacturing industry, specifically an adaptive multi-depth-of-field sub-pixel layer defect detection method, which is a specific application of traditional machine vision algorithms in industrial manufacturing, and is particularly focused on the detection needs of the sub-pixel layer. With the rapid development of electronic display technology, especially the widespread application of high-resolution LCD and OLED displays, the requirements for screen display quality are becoming increasingly higher. In the production process of display screens, defect detection at the sub-pixel layer is a key link in ensuring product quality and improving production efficiency. Background Art

[0002] Currently, traditional defect detection methods primarily rely on single-depth-of-field photography. This involves capturing images of the display using a camera with a fixed depth of field and then using machine vision algorithms to detect defects. However, this approach has significant limitations. Due to the complex physical structure of the sub-pixel layer, defects may appear at different depths of field. Single-depth-of-field photography struggles to capture all defects, leading to incomplete detection or false detections. Furthermore, existing algorithms lack targeted optimization for images at different depths of field, limiting detection accuracy and efficiency.

[0003] To overcome these technical challenges, the present inventors explored multi-depth-of-field (MDF) imaging technology. By capturing images at different depths, more comprehensive defect information can be captured. However, effectively adapting and processing multi-depth images remains an urgent challenge. Existing multi-depth inspection solutions often use a unified algorithm to process all depth images, failing to fully utilize the unique information at each depth, resulting in suboptimal inspection results. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides an adaptive multi-depth sub-pixel layer defect detection method, which can automatically identify and determine the clearest layer position of the defect in the sub-pixel layer through multi-depth image acquisition and adaptive detection algorithm, thereby achieving accurate positioning and efficient detection of the defect. The present invention significantly improves the accuracy and efficiency of defect detection, and provides an innovative solution for product quality control and production efficiency improvement in the electronic screen manufacturing industry; in actual production, this technology can be widely used in the manufacturing process of electronic screens, including but not limited to sub-pixel layer defect detection of LCD, OLED and other display screens. Through comprehensive and accurate defect detection, this technology can effectively improve product quality, reduce defective rates, optimize production processes, and reduce production costs. Therefore, the present invention has important application value in the electronic screen manufacturing industry, and provides strong technical support for improving the intelligence and automation level of industrial manufacturing.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An adaptive multi-depth sub-pixel layer defect detection method comprises the following steps:

[0007] Step S1: Capture images of the sub-pixel layer through an image acquisition module to obtain images of different depths of field;

[0008] Step S2: pre-processing the collected image based on the image pre-processing module, specifically using multiple filters to reduce noise and smooth the input image to eliminate noise interference and improve image quality;

[0009] Step S3: using the defect detection module to accurately extract edge features in images of different depths of field;

[0010] Step S4: Calculating the multi-dimensional feature information of the image through the feature calculation module;

[0011] Step S5: Based on the multi-dimensional feature information obtained above, defect evaluation indicators for human visual perception are designed. By adaptively adjusting the weights of each evaluation indicator and combining the depth information of the multi-depth-of-field image, the clearest layer where the defect is located is accurately located, thereby determining the optimal distribution position of the defect in the sub-pixel layer. At the same time, a comparative analysis mechanism consistent with human visual perception is introduced to ensure that the detection results are highly consistent with human visual perception.

[0012] As a further solution of the present invention, in step S1, the image acquisition module uses a 12-megapixel black and white camera configured with a telecentric lens to capture pictures with five different depths of field at the sub-pixel layer.

[0013] As a further solution of the present invention, in step S2: specifically, by introducing a Fourier transform filter, a guided filter, a Gaussian filter, a median filter and a bilateral filter, the optimal filter combination is adaptively selected to effectively retain image details while removing noise.

[0014] As a further solution of the present invention, in step S3: a multi-scale edge detection algorithm is used, combined with the Canny operator, Sobel operator, Prewitt operator and Laplacian operator to extract edge features in images of different depths of field, wherein the Canny operator is used to detect high-contrast edges, the Sobel operator and Prewitt operator enhance the directionality of the edge, and the Laplacian operator is used to detect the sharpness of the edge, and after applying a Gaussian filter to smooth the image, the horizontal and vertical differences are calculated to enhance the directionality and clarity of the edge.

[0015] As a further solution of the present invention, in step S4, the multi-dimensional feature information of the image includes standard deviation, gradient amplitude, and Laplace response, and the texture complexity of the image area is characterized by the standard deviation: ,in, N is the total amount of data, μ is the overall mean, It is i data points; the gradient magnitude reflects the strength of the image edge:

[0016] Horizontal gradient: ;

[0017] Vertical Gradient: ;

[0018] Gradient Magnitude: ;

[0019] in, Gx : Indicates the horizontal gradient of the image, reflecting the rate of change of the brightness of the image in the horizontal direction. Gy : Indicates the vertical gradient of the image, reflecting the rate of change of the brightness of the image in the vertical direction. G : represents the gradient amplitude, that is, the length of the gradient vector, which comprehensively reflects the degree of brightness change of the image in the horizontal and vertical directions at that point;

[0020] The Laplace response is used to detect the sharpness of the image

[0021] in: L ( x , y ): indicates the coordinate (x,y) ( x , y ) is used to measure the edge strength of the point;

[0022] f(x+1,y): represents the value at x x Gray value after shifting one pixel to the right in the direction;

[0023] f(x−1,y): represents the x Gray value after shifting one pixel to the left in the direction;

[0024] f(x,y+1): represents the value at y y The grayscale value after moving up one pixel in the direction;

[0025] f(x,y−1): represents the y The grayscale value after moving down one pixel in the direction;

[0026] 4f(x,y): represents the current pixel (x,y) ( x , y ) is four times the grayscale value at ).

[0027] As a further solution of the present invention, in step S5, the defect evaluation index of human visual perception is designed based on the obtained multidimensional feature information, specifically: the edge sharpness is evaluated by calculating the gradient amplitude and Laplace response of the edge. The larger the gradient amplitude and Laplace value, the more dramatic the brightness change near the point, and the edge is more obvious and sharper. Conversely, the brightness change near the point is gentle and the edge is not obvious enough. The standard deviation is used to measure the distribution of pixel values ​​in the image. A large standard deviation indicates that the pixel values ​​in the image change greatly, indicating that there are more details or textures in the image. A small standard deviation indicates that the pixel values ​​change more slowly, and the image appears more uniform or blurred.

[0028] As a further solution of the present invention, in step S5, by adaptively adjusting the weights of each evaluation index and combining the depth information of the multi-depth image, the clearest layer where the defect is located is accurately located, thereby determining the optimal distribution position of the defect in the sub-pixel layer. At the same time, a comparative analysis mechanism consistent with human visual perception is introduced to ensure that the detection results are highly consistent with human visual perception. The specific process includes: proposing a weight adjustment formula based on human eye perception: .in, It is i The weight of the evaluation indicators, α and β is an adaptive parameter, S i It is the multi-dimensional feature information extracted by the feature calculation module. D i It is the depth information of the multi-depth image.

[0029] The present invention also provides an adaptive multi-depth sub-pixel layer defect detection device, including an image acquisition module, which is used to acquire pictures of the sub-pixel layer to obtain images of different depths of field; an image preprocessing module, which uses a variety of filters to reduce noise and smooth the input image to eliminate noise interference and improve image quality; a defect detection module, which is used to accurately extract edge features in images of different depths of field; a feature calculation module, which is used to calculate the multi-dimensional feature information of the image; an adaptive evaluation algorithm module, which adaptively adjusts the weights of each evaluation indicator and combines the depth information of the multi-depth images to accurately locate the clearest layer where the defect is located, thereby determining the optimal distribution position of the defect in the sub-pixel layer, and at the same time introduces a comparative analysis mechanism consistent with the human eye's visual perception to ensure that the detection results are highly consistent with human visual perception.

[0030] The present invention has the following beneficial effects:

[0031] 1) Improved Defect Detection Accuracy: This invention utilizes multi-depth-of-field imaging technology combined with an adaptive algorithm to comprehensively capture defect details at the sub-pixel level. It also uses an adaptive evaluation algorithm to precisely locate the clearest layer where the defect is located. This method, based on human visual perception, ensures that detection results are highly consistent with human visual perception, significantly improving the accuracy and reliability of defect detection.

[0032] 2) Improved Inspection Efficiency: This invention utilizes multi-depth-of-field photography and an adaptive detection algorithm to achieve rapid and accurate inspection of sub-pixel images. Compared to traditional single-depth-of-field photography methods, this technology automates defect detection and assessment, significantly reducing the need for manual inspection and significantly improving inspection efficiency. This technology provides effective technical support for product quality control in the electronic display manufacturing industry.

[0033] 3) Enhanced robustness and adaptability: The adaptive algorithm in this paper automatically adjusts detection parameters and evaluation metric weights based on image depth information and feature changes, adapting to varying depths of field and image conditions. This method is not only suitable for standard sub-pixel defect detection but also demonstrates strong adaptability across diverse manufacturing environments and product types, demonstrating broad applicability and robustness.

[0034] In order to more clearly illustrate the structural features and effects of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic diagram of the camera layered shooting mentioned in the present invention.

[0036] Figure 2a-2e For Figure 1 Schematic diagram of the imaging effects of the middle CG layer (color filter layer), DITO layer (double-layer ITO conductive layer), CF layer (common electrode layer), Bottom Pol layer (lower polarizing layer) and Top Diffuser layer (diffusion layer) at different depths of field.

[0037] Figure 3 It is the overall algorithm flow chart mentioned in the present invention.

[0038] Figure 4a-4c This is the image preprocessing process diagram mentioned in the present invention.

[0039] Figure 5a The following figure shows the original image of layer 1.4 (Top Diffuser layer), including schematic diagrams of various potential defects.

[0040] Figure 5b To show the extracted Figure 5a Schematic diagram of defect outline.

[0041] Figure 6 This is a schematic diagram comparing the clearest layer mentioned in the present invention and the AI ​​visual area. DETAILED DESCRIPTION

[0042] The present invention will be further described below with reference to the accompanying drawings and related knowledge, and described clearly and completely. Obviously, the described applications are only part of the embodiments of the present invention, rather than all of the embodiments.

[0043] The present invention provides an adaptive multi-depth sub-pixel layer defect detection method. Through multi-depth image acquisition and adaptive detection algorithms, it can automatically identify and determine the clearest layer position of defects in the sub-pixel layer, thereby achieving accurate positioning and efficient detection of defects. The present invention significantly improves the accuracy and efficiency of defect detection, and provides an innovative solution for product quality control and production efficiency improvement in the electronic screen manufacturing industry. The core idea of ​​the present invention is to achieve comprehensive and accurate detection of sub-pixel layer images through multi-depth shooting technology combined with adaptive detection algorithms.

[0044] The present invention provides an adaptive multi-depth sub-pixel layer defect detection method, comprising the following steps:

[0045] Step S1: The image acquisition module is used to acquire images of the sub-pixel layer to obtain images of different depths of field; specifically, a 12 million black and white camera is used with a telecentric lens to capture images of 5 different depths of field at the sub-pixel layer (e.g. Figure 1 As shown), image effects (such as Figure 2a-2e (as shown) covers different depth areas, thus ensuring comprehensive capture of defect details; Figure 1 In the figure, mark 1 indicates the first layer shooting CG layer, mark 2 indicates the second layer shooting DITO layer, mark 3 indicates the third layer shooting CF layer, mark 4 indicates the fourth layer shooting Bottom POL layer, mark 5 indicates the fifth layer shooting Top Diffuser layer, mark 6 is the upper polarizer + color filter layer, mark 7 is the thin film transistor + lower polarizer layer. And in Figure 2a This is a schematic diagram of the CG shooting effect. Figure 2b This is a schematic diagram of the DITO shooting effect. Figure 2c This is a schematic diagram of the CF shooting effect. Figure 2d This is a diagram of the shooting effect of Bottom Pol. Figure 2e Schematic diagram of the shooting effect of Top Diffuser.

[0046] In this invention, a 12-megapixel black-and-white camera paired with a telecentric lens is used. This lens reduces image distortion caused by viewing angles, ensuring high-precision and accuracy in captured images, providing high-quality raw data for subsequent analysis. Furthermore, images are captured at five different depths of field at the sub-pixel level. By setting these depths of field, the captured images cover different depths of the sub-pixel layer, fully capturing potential defect details and avoiding missing defects at different depths.

[0047] Step S2: Preprocess the captured image using the image preprocessing module. Specifically, it applies multiple filters to reduce noise and smooth the input image to eliminate noise interference and improve image quality. This step is specifically the adaptive detection algorithm processing stage (this module is the core of this solution). First, in the image preprocessing module, multiple filters are applied to reduce noise and smooth the input image to eliminate noise interference and improve image quality. This module introduces Fourier transform filters, guided filters, Gaussian filters, median filters, and bilateral filters. By adaptively selecting the optimal filter combination, it can effectively preserve image details while removing noise interference, improving image quality and providing a clear image foundation for subsequent defect detection. This module is the core component of image preprocessing. By adaptively selecting filter combinations, it can automatically adjust filter parameters and weights based on the characteristics of different images and noise distribution to achieve optimal noise reduction and smoothing effects.

[0048] Step S3: Accurately extract edge features from images at different depths of field using the defect detection module. Specifically, the defect detection module uses a multi-scale edge detection algorithm, combined with multiple edge detection operators such as the Canny, Sobel, Prewitt, and Laplacian operators, to extract edge features from the image. This module adaptively adjusts edge detection parameters and incorporates sub-pixel depth information to accurately extract edge features from images at different depths of field, providing high-quality edge information for subsequent defect detection.

[0049] By combining multiple operators, the present invention enables more comprehensive extraction of edge features in images. Furthermore, the module adaptively adjusts edge detection parameters and, in combination with sub-pixel depth information, automatically adjusts operator parameters, such as thresholds and convolution kernel sizes, based on the characteristics and edge distribution of images at different depths of field. This allows for precise edge feature extraction in images at varying depths of field. This provides high-quality edge information for subsequent defect detection, accurately locating defect edges.

[0050] Step S4: Calculate the multidimensional feature information of the image through the feature calculation module. Specifically, in the feature calculation module, extract the multidimensional feature information of the image, including standard deviation, gradient amplitude, Laplacian response, texture features, etc. These features can comprehensively characterize the local characteristics of the image and provide rich criteria for defect detection.

[0051] Specifically, the feature calculation module extracts multidimensional image feature information, including standard deviation, gradient magnitude, Laplacian response, and texture features. The standard deviation reflects the degree of dispersion of the image's grayscale values, the gradient magnitude indicates the rate of change of pixels within the image, the Laplacian response is sensitive to features such as edges and corners, and texture features describe the image's local texture structure. These features comprehensively characterize the image's local characteristics, providing a rich set of criteria for defect detection. By calculating these multidimensional features, defect characteristics can be characterized from various perspectives. For example, a defect may cause significant local grayscale variations, thereby increasing the standard deviation and gradient magnitude; or it may alter the image's texture structure, which can be detected using texture features.

[0052] Step S5: Based on the multi-dimensional feature information obtained above, defect assessment indicators based on human visual perception are designed. By adaptively adjusting the weights of each evaluation indicator and combining the depth information of the multi-depth images, the clearest layer where the defect is located is accurately located, thereby determining the optimal distribution position of the defect in the sub-pixel layer. A comparative analysis mechanism consistent with human visual perception is also introduced to ensure that the detection results are highly consistent with human visual perception. Specifically, within the adaptive evaluation algorithm system, a set of defect assessment indicators based on human visual perception is designed, focusing on key factors such as edge sharpness, clarity, contrast, and uniformity. By adaptively adjusting the weights of each evaluation indicator and combining the depth information of the multi-depth images, this evaluation system can accurately locate the clearest layer where the defect is located, thereby determining the optimal distribution position of the defect in the sub-pixel layer. Furthermore, the evaluation algorithm introduces a comparative analysis mechanism consistent with human visual perception to ensure that the detection results are highly consistent with human visual perception, thereby improving the reliability and practicality of the detection results. In summary, the proposed method significantly improves the accuracy and efficiency of defect detection, providing reliable technical support for quality control in the electronic screen manufacturing industry.

[0053] In this paper, a set of defect assessment metrics based on human visual perception is designed within the adaptive assessment algorithm system, focusing on key factors such as edge sharpness, clarity, contrast, and uniformity. Edge sharpness reflects the clarity of the defect edge, clarity indicates the overall clarity of the image, contrast reflects the grayscale differences between different areas in the image, and uniformity describes the uniformity of the image's grayscale distribution. By comprehensively considering these factors, the severity and location of defects can be more accurately assessed.

[0054] This evaluation system adaptively adjusts the weights of each evaluation metric, incorporating the depth information of multi-depth images. It automatically adjusts the importance of each metric based on the characteristics of each image and the distribution of defects. For example, for defects with distinct edge features, edge sharpness may be given a higher weight; whereas for defects that affect overall image clarity, clarity may be given a higher weight. This allows precise location of the sharpest layer where the defect resides, thereby determining the optimal distribution of defects within the sub-pixel layer.

[0055] Furthermore, the evaluation algorithm incorporates a comparative analysis mechanism for consistency with human visual perception, comparing and analyzing test results with human visual perception. By simulating the human eye's image perception process, detected defects are further verified and adjusted to ensure high consistency between test results and human visual perception. This improves the reliability and practicality of test results, making them more realistic and easier for operators to judge and process.

[0056] The method of the present invention significantly improves the accuracy and efficiency of sub-pixel layer defect detection on electronic screens through a series of steps including image acquisition, preprocessing, edge feature extraction, feature calculation, and defect assessment and positioning, providing reliable technical support for quality control in the electronic screen manufacturing industry.

[0057] Further preferably, refer to Figure 4a-4c As shown, the image preprocessing module, the main purpose of the image preprocessing module is to reduce noise and smooth the input multi-depth images while retaining the image detail information. This module uses a variety of filters to process the image, including but not limited to the following: Fourier transform filter, guided filter, Gaussian filter, median filter and bilateral filter, etc. In the present invention, the adaptive image preprocessing module uses an adaptive combination of multiple filters (such as Fourier transform filter, guided filter, Gaussian filter, etc.) in the image preprocessing stage to effectively remove noise and improve image quality, while retaining image detail information, providing high-quality input for subsequent defect detection. Figure 4a As shown, this is the original image of the Top Diffuser layer. Figure 4b For process diagram, Figure 4c The preprocessed image.

[0058] In the present invention, the Fourier transform filter can convert an image from the spatial domain to the frequency domain. In the frequency domain, it can easily distinguish between the low-frequency components (such as the overall structure and shape) and high-frequency components (such as edges, textures, and noise) of the image. By designing different frequency domain filters, such as low-pass filters that can filter out high-frequency noise and high-pass filters that can enhance details such as edges, image noise reduction and detail enhancement can be achieved. The guided filter is an edge-preserving filter that can use a guiding image to guide the filtering process. While smoothing the image, it can well preserve the edge and detail information of the image. It is particularly suitable for preprocessing images with complex textures and edge structures. The Gaussian filter uses a Gaussian function to filter the image, which can effectively remove Gaussian noise in the image, smooth the image, and to a certain extent preserve the edge information of the image, because the characteristics of the Gaussian function make it possible to smooth the noise without excessively blurring the details of the image. The median filter is a nonlinear filter that replaces the value of each pixel with the median of its neighboring pixel values. This filter is very effective for removing impulsive noise such as salt and pepper noise, while also preserving image edges and sharpness, avoiding the edge blurring that can occur with linear filters during the denoising process. A bilateral filter combines the spatial proximity of an image with the similarity of pixel values ​​for filtering. It considers not only the spatial distance between pixels but also the differences in pixel values. This allows it to effectively preserve image edges and details while removing noise, preventing over-smoothing of image edges.

[0059] The present invention takes advantage of the adaptive combination of multiple filters: different filters have different processing effects on different types of noise and image details. For example, a Gaussian filter works well for Gaussian noise, but may blur details; a median filter is effective for salt and pepper noise and can preserve edges, but has weaker processing capabilities for Gaussian noise. By adaptively combining these filters, the most appropriate filter or filter combination can be automatically selected for processing based on the specific characteristics and noise conditions of the image, thereby effectively removing noise while retaining the image's detail information to the greatest extent possible, providing high-quality input images for subsequent defect detection. For example, for images with more noise and rich edge details, a combination of a median filter and a bilateral filter may be adaptively selected; for images containing Gaussian noise and needing to retain a certain amount of low-frequency information, a combination of a Gaussian filter and a Fourier transform filter may be selected, etc.

[0060] Further preferably, refer to Figure 5a and Figure 5bAs shown in Figure 2, the goal of the multi-scale defect edge detection module is to extract edge features in the image for subsequent defect detection. This algorithm is implemented by combining multiple edge detection operators and filter difference technology.

[0061] The following is a detailed description of this algorithm and its advantages: It applies the Canny, Sobel, Prewitt, and Laplacian operators at different scales. The Canny operator is used to detect high-contrast edges, the Sobel and Prewitt operators enhance edge directionality, and the Laplacian operator detects edge sharpness. Through multi-scale processing, it captures edge features of varying sizes and orientations. The filtered difference technique, after smoothing the image with a Gaussian filter, calculates horizontal and vertical differences to enhance edge directionality and clarity. Furthermore, a bilateral filter is used for further smoothing while preserving edge details. The advantages of this method are: 1) Complementarity: The Canny operator performs well in images with less noise, while the Sobel and Prewitt operators are more robust in images with more noise. The Laplacian operator is sensitive to edge sharpness and is suitable for detecting high-contrast edges. Combining these operators comprehensively captures different types of edge features. 2) Accuracy: The combination of multiple operators improves edge detection accuracy and reduces the errors that may occur when using a single operator under specific conditions. 3) Robustness: Multi-scale processing can maintain detection stability under different depth of field and lighting conditions.

[0062] It should be noted in the present invention that different operators have their own advantages under different image conditions. When used in combination, they can comprehensively capture various types of edge features and better detect edges regardless of the image noise level; multiple operators complement and verify each other, reducing the errors caused by a single operator under specific conditions due to its own limitations, thereby improving the accuracy of edge detection and making the detection results closer to the real edge; and multi-scale processing can capture more stable edge features by analyzing images at different scales under different depths of field and lighting conditions, reducing the impact of environmental changes on detection results and maintaining detection stability.

[0063] In the present invention, the process of calculating the multidimensional feature information of an image through the feature calculation module specifically includes: the main task of the feature calculation module is to extract the multidimensional feature information of the image, including standard deviation, gradient amplitude, Laplace response, etc.

[0064] 1) Standard deviation characterizes the texture complexity of the image area:

[0065] in, N is the total amount of data, μ is the overall mean, It is i data points.

[0066] 2) Gradient amplitude reflects the strength of the image edge

[0067] Horizontal gradient:

[0068] Vertical Gradient:

[0069] Gradient Magnitude:

[0070] in, Gx : Indicates the gradient of the image in the horizontal direction (x-axis direction), reflecting the rate of change of the brightness of the image in the horizontal direction. Gy : Indicates the gradient of the image in the vertical direction (y-axis direction), reflecting the rate of change of the brightness of the image in the vertical direction. G : Indicates the gradient amplitude, that is, the length of the gradient vector, which comprehensively reflects the degree of brightness change of the image in the horizontal and vertical directions at that point.

[0071] 3) Laplace response is used to detect the sharpness of the image

[0072] in: L ( x , y ): indicates the coordinate (x,y) ( x , y ), which is used to measure the edge strength at that point.

[0073] f(x+1,y): represents the value at x x The grayscale value after shifting one pixel to the right in the vertical direction.

[0074] f(x−1,y): represents the x The grayscale value after shifting one pixel to the left in the vertical direction.

[0075] f(x,y+1): represents the value at y y The grayscale value after moving up one pixel in the direction.

[0076] f(x,y−1): represents the y The grayscale value after moving down one pixel in the direction.

[0077] 4f(x,y): represents the current pixel (x,y) ( x , y ) is four times the grayscale value at ).

[0078] In this invention, the adaptive evaluation algorithm module is the core of the entire method, used to evaluate and classify defects at different sub-pixel levels. This module designs a set of defect assessment indicators based on human visual perception, focusing on factors such as edge sharpness and image contrast. Edge sharpness is assessed by calculating the gradient amplitude and Laplace response of the edge. Larger gradient amplitudes and Laplace values ​​indicate a more dramatic brightness change near that point, resulting in a more distinct and sharp edge. Conversely, a more gradual brightness change near that point indicates a less distinct edge. The standard deviation can be used to measure the distribution of pixel values ​​in an image. A large standard deviation indicates a greater variation in pixel values ​​within the image, indicating more detail or texture. A small standard deviation indicates a more gradual variation in pixel values, resulting in a more uniform or blurred image. The multidimensional feature extraction and adaptive evaluation system in this invention extracts multidimensional feature information from the image (such as standard deviation, gradient amplitude, Laplace response, texture features, etc.) and designs a set of adaptive evaluation algorithms based on human visual perception. The algorithm adaptively adjusts the weights of each evaluation metric and, in combination with the depth information of multi-depth images, accurately locates the clearest layer where defects are located, ensuring that the detection results are highly consistent with human visual perception. Based on the adaptive evaluation mechanism of human visual perception, this invention introduces a comparative analysis mechanism consistent with human visual perception during the defect assessment phase, focusing on key factors such as edge sharpness, clarity, contrast, and uniformity to ensure the reliability and practicality of the detection results.

[0079] Furthermore, the algorithm adaptively adjusts the weights of various evaluation metrics and, in combination with the depth information of multi-depth images, accurately locates the clearest layer where defects are located. This means that the optimal distribution of defects within the sub-pixel layer can be accurately determined, avoiding misjudgments or omissions of defects due to varying depths of field. Furthermore, a comparative analysis mechanism consistent with human visual perception is introduced, fully considering key factors such as edge sharpness, clarity, contrast, and uniformity. This ensures that the detection results are highly consistent with human visual perception, thereby improving their reliability and practicality and better meeting the requirements for defect detection in real-world application scenarios.

[0080] In the present invention, reference is made to Figure 6 As shown in the figure, the analysis diagram of the clearest layer selected by this module and the consistency with AI is shown. Specifically, the evaluation algorithm achieves accurate identification and classification of defects in different sub-pixel layers by adaptively adjusting the weights of each evaluation indicator and combining the depth information of multiple depth-of-field images. First, the weight adjustment formula based on human eye perception is proposed as follows: .in, It is i The weight of the evaluation indicators, α and β is an adaptive parameter, S iIt is the characteristic value extracted by the characteristic calculation module (such as standard deviation, gradient amplitude, Laplacian response, etc.), D i This is the depth information for the multi-depth image. Its values ​​are 0.2, 0.5, 0.8, 1.1, and 1.4, corresponding to the CG layer, DITO layer, CF layer, Bottom Pol layer, and TopDiffuser layer, respectively. Specifically, when the image has high edge sharpness (i.e., large gradient amplitude) and high image contrast (i.e., large standard deviation) in a particular layer, the weight of that layer will also be larger. By comparing the weights of the five layers, we can determine the optimal distribution of defects in the sub-pixel layer and determine which depth layer is clearest.

[0081] This method combines multi-depth photography with an adaptive algorithm to capture sub-pixel images at different depths of field. This technology, combined with an adaptive algorithm, enables precise defect location and detection. This method automatically determines the location of defects in the clearest layer, significantly improving detection accuracy and efficiency.

[0082] In addition, in the present invention, multi-scale edge detection is combined with depth information. In the defect detection stage, a multi-scale edge detection algorithm is used in combination with a variety of edge detection operators (such as Canny, Sobel, Prewitt, etc.), and depth information is introduced to adaptively adjust the detection parameters to achieve accurate extraction of edge features in images with different depths of field.

[0083] The following provides specific embodiments of the present invention.

[0084] Example 1, a method for detecting sub-pixel layer defects with adaptive multi-depth of field, referring to Figure 3 The overall algorithm flow chart shown in FIG. Specifically, the sub-pixel layer image is collected by the image acquisition module to obtain images of different depths of field. Figure 1 The figure shows a schematic diagram of a camera capturing sub-pixel layers of an electronic screen using multiple depths of field. The diagram shows the imaging effects of the CG layer (color filter layer), DITO layer (double-layer ITO conductive layer), CF layer (common electrode layer), Bottom Pol layer (bottom polarizing layer), and Top Diffuser layer (diffusion layer) at different depths of field. Multi-depth-of-field capture technology clearly captures the optical characteristics and potential defects of each layer. Specifically, different depths of field correspond to the depth positions of different sub-pixel layers, fully covering all key levels of the screen structure. This diagram intuitively demonstrates the application of multi-depth-of-field capture in sub-pixel layer defect detection, providing multi-dimensional image data support for subsequent adaptive algorithms, thereby achieving precise defect location and classification. This multi-depth-of-field layered capture method effectively addresses the limitations of traditional single-depth-of-field detection methods, significantly improving the comprehensiveness and accuracy of defect detection.

[0085] The collected images are preprocessed based on the image preprocessing module. Specifically, a variety of filters are used to reduce noise and smooth the input images to eliminate noise interference and improve image quality. Figure 2a-2e The figure shows the product's imaging results under five depths of field, visually demonstrating the optical characteristics and image quality of the sub-pixel layer at different depths. Using multi-depth imaging technology, the imaging differences of the CG, DITO, CF, Bottom Pol, and Top Diffuser layers at different depths can be clearly observed. The imaging results of each layer reflect its optical properties and potential defects, such as the clarity of the CG layer, the edge characteristics of the DITO layer, the color uniformity of the CF layer, the structural integrity of the Bottom Pol layer, and the light diffusion characteristics of the Top Diffuser layer. These multi-depth images provide rich image data support for the subsequent adaptive defect detection algorithm, enabling the algorithm to accurately locate and classify defects in the sub-pixel layer by combining depth information and image features. This figure intuitively demonstrates the application value of multi-depth imaging technology in sub-pixel layer defect detection, providing strong technical support for quality control in the electronic display manufacturing industry.

[0086] Figure 4 shows a schematic diagram of the image preprocessing process, illustrating the combined application of multiple filters and their effects on the image. During the image preprocessing stage, the system applies a variety of filters (such as Fourier transform filters, guided filters, Gaussian filters, median filters, and bilateral filters), adaptively selecting the optimal filter combination to reduce noise and smooth the input multi-depth images. Specifically, to address the differences in the characteristics of images at different depths, we first analyze the noise type, texture features, and edge information of each image layer. For example, the Top Diffuser layer is primarily affected by complex background textures, so a combination of a median filter and a guided filter is used to effectively remove background texture and reduce noise. The CF layer, which has complex background textures and Gaussian noise, uses a combination of a Gaussian filter and a bilateral filter to effectively remove noise while preserving image details. This provides a high-quality image foundation for subsequent defect detection.

[0087] These filters work together to effectively remove noise interference from the image while retaining the image's detailed information, thereby improving the overall image quality and the accuracy of subsequent defect detection. Figure 4a-4c The application effects of different filters in the preprocessing process are intuitively demonstrated, reflecting the important role of the preprocessing module in optimizing image quality and providing high-quality input for subsequent defect detection.

[0088] Reference Figure 5a 、 Figure 5b As shown in Figure 2, the results of defect extraction are shown, where Figure 5a The original image of layer 1.4 is shown, including various potential defects. Figure 5b The algorithm-extracted defect outline is clearly displayed. The method of the present invention utilizes multi-depth-of-field imaging technology to acquire image information at different depths of field, and combines it with an adaptive algorithm to perform image preprocessing, edge detection, and feature extraction. Through these steps, the system can accurately identify and locate defects, capturing their detailed outlines.

[0089] Reference Figure 6 As shown in the figure, the comparison results of the clearest layer and the AI ​​visual area are displayed. The 10 figures intuitively show the high consistency between the defect detection results of the algorithm of the present invention and the results of manual visual inspection, proving the reliability of the algorithm of the present invention in defect detection accuracy.

[0090] The technical principles of the present invention have been described above in conjunction with specific embodiments, which are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention fall within the scope of protection of the present invention. Those skilled in the art will be able to conceive of other specific embodiments of the present invention without inventive effort, and such methods will fall within the scope of protection of the present invention.

Claims

1. An adaptive multi-depth sub-pixel layer defect detection method, characterized in that: The following steps are involved: Step S1: Capture images of the sub-pixel layer through an image acquisition module to obtain images of different depths of field; Step S2: Preprocessing the collected images of different depths of field based on the image preprocessing module, specifically using multiple filters to reduce noise and smooth the input images of different depths of field to eliminate noise interference and improve image quality. Specifically, by introducing Fourier transform filters, guided filters, Gaussian filters, median filters and bilateral filters, the optimal filter combination is adaptively selected to achieve the goal of retaining image details while removing noise; Step S3: using a defect detection module to accurately extract edge features from pre-processed images of different depths of field; Step S4: Calculating multi-dimensional feature information of the image through a feature calculation module based on the extracted edge features; Step S5: Based on the multi-dimensional feature information obtained above, a defect evaluation index of human visual perception is designed. By adaptively adjusting the weight of each evaluation index and combining the depth information of the multi-depth image, the clearest layer where the defect is located is accurately located, thereby determining the optimal distribution position of the defect in the sub-pixel layer. At the same time, a comparative analysis mechanism consistent with human visual perception is introduced to ensure that the detection results are highly consistent with human visual perception. By adaptively adjusting the weight of each evaluation index and combining the depth information of the multi-depth image, the clearest layer where the defect is located is accurately located, thereby determining the optimal distribution position of the defect in the sub-pixel layer. At the same time, a comparative analysis mechanism consistent with human visual perception is introduced to ensure that the detection results are highly consistent with human visual perception. The specific process includes: proposing a weight adjustment formula based on human eye perception as follows: ,in, is the weight of the i-th evaluation index, α and β are adaptive parameters, Si is the multi-dimensional feature information extracted by the feature calculation module, and Di is the depth information of the multi-depth image; through multi-depth shooting and adaptive detection algorithm, fast and accurate detection of sub-pixel layer images is achieved, and the adaptive algorithm automatically adjusts the detection parameters and evaluation index weights according to the depth information and feature changes of the image to adapt to changes in different depths of field and image conditions.

2. The method for detecting sub-pixel defects with adaptive multi-depth of field according to claim 1, wherein: In step S1, the image acquisition module uses a 12-megapixel black-and-white camera configured with a telecentric lens to capture images with five different depths of field at the sub-pixel layer.

3. The method for detecting sub-pixel defects with adaptive multi-depth of field according to claim 2, wherein: In step S3: a multi-scale edge detection algorithm is used, combining the Canny operator, Sobel operator, Prewitt operator and Laplacian operator to extract edge features in images of different depths of field. The Canny operator is used to detect high-contrast edges, the Sobel operator and Prewitt operator enhance the directionality of the edge, and the Laplacian operator is used to detect the sharpness of the edge. After applying a Gaussian filter to smooth the image, the horizontal and vertical differences are calculated to enhance the directionality and clarity of the edge.

4. The method for detecting sub-pixel defects with adaptive multi-depth of field according to claim 3, wherein: In step S4, the multi-dimensional feature information of the image includes standard deviation, gradient amplitude, and Laplace response, and the texture complexity of the image area is characterized by the standard deviation: ,in, N is the total amount of data, μ is the overall mean, It is i data points; the gradient magnitude reflects the strength of the image edge: Horizontal gradient: ; Vertical Gradient: ; Gradient Magnitude: ; in, Gx : Indicates the horizontal gradient of the image, reflecting the rate of change of the brightness of the image in the horizontal direction. Gy : Indicates the vertical gradient of the image, reflecting the rate of change of the brightness of the image in the vertical direction. G : represents the gradient amplitude, that is, the length of the gradient vector, which comprehensively reflects the degree of brightness change of the image in the horizontal and vertical directions at that point; The Laplace response is used to detect the sharpness of the image; ； in: L ( x , y ): represents the Laplace response value at the coordinate (x, y), which is used to measure the edge strength of the point; f(x+1,y): represents the grayscale value after shifting one pixel to the right in the x direction; f(x-1,y): represents the grayscale value after shifting one pixel to the left in the x direction; f(x,y+1): represents the grayscale value after moving up one pixel in the y direction; f(x,y-1): represents the grayscale value after moving down one pixel in the y direction; 4f(x,y): represents four times the grayscale value of the current pixel (x,y).

5. The method for detecting sub-pixel defects with adaptive multi-depth of field according to claim 4, wherein: In step S5, a defect assessment index of human visual perception is designed based on the obtained multidimensional feature information, specifically: edge sharpness is evaluated by calculating the gradient amplitude and Laplace response of the edge. The larger the gradient amplitude and Laplace value, the more dramatic the brightness change near the point, and the edge is more obvious and sharper. Conversely, the brightness change near the point is gentle, and the edge is not obvious enough. The standard deviation is used to measure the distribution of pixel values ​​in the image. A large standard deviation indicates a large change in pixel values ​​in the image, indicating that there are more details or textures in the image. A small standard deviation indicates a gentle change in pixel values, and the image appears uniform or blurred.

6. A detection device using the adaptive multi-depth sub-pixel layer defect detection method according to claim 1, characterized in that: It includes an image acquisition module, which is used to collect pictures at the sub-pixel layer to obtain images with different depths of field; an image preprocessing module, which uses multiple filters to reduce noise and smooth the input image to eliminate noise interference and improve image quality; Defect detection module, used to accurately extract edge features in images of different depths of field; Feature calculation module, used to calculate the multi-dimensional feature information of the image; The adaptive evaluation algorithm module adaptively adjusts the weights of each evaluation indicator and combines the depth information of multi-depth images to accurately locate the clearest layer where the defect is located, thereby determining the optimal distribution position of the defect in the sub-pixel layer. At the same time, it introduces a comparative analysis mechanism consistent with human visual perception to ensure that the detection results are highly consistent with human visual perception.

Images