A method for detecting the image of an ultra-thin narrow-frame touch-sensitive liquid crystal display screen

Abstract

The present invention relates to the field of image processing, and in particular to a method for detecting images on an ultra-thin narrow-frame touch-screen LCD screen. The content includes: constructing a hybrid progressive dynamic learning network by combining high-order dynamic recursive learning with hybrid feature adaptive enhancement technology; and enhancing the feature map obtained by the high-order dynamic recursive learning model using an adaptive feature enhancement strategy; introducing an adaptive lighting adjustment algorithm to dynamically adjust the brightness and contrast of the image; and further using a dynamic content adjustment algorithm based on image processing to identify and process defective areas in the image. The method solves the technical problems that the prior art lacks high-sensitivity recognition of complex image backgrounds and tiny defect features, making the detection of defects such as dead spots, bright spots and color differences in edge and corner areas inaccurate and incomplete; fails to effectively adapt to different ambient lighting conditions; and lacks dynamic parameter adjustment and feature enhancement mechanisms.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen. Background Art

[0002] In the current consumer electronics market, with the rapid development of mobile devices and display technology, users' requirements for the quality and aesthetics of display screens are constantly increasing. The wide application of ultra-thin narrow-frame touch LCD screens in smartphones, tablets, laptops and other portable devices has put forward higher standards for display clarity, color authenticity and overall visual experience. However, due to technical limitations and material defects in the production process, LCD screens often have tiny defects such as dead spots, bright spots, and color differences, which greatly affect the overall quality of the product and user experience.

[0003] In this context, although a variety of defect detection technologies have been proposed and applied in production and quality control processes, these technologies often find it difficult to effectively identify and process tiny defects in ultra-thin narrow bezel environments, and the accuracy and stability of defect detection under different ambient lighting conditions remains a challenge.

[0004] my country's patent application number: CN202311227287.7, publication date: 2023.11.28, discloses a method and device for detecting defects in liquid crystal displays based on deep learning, which relates to the field of computer vision target detection technology. By designing a lightweight feature extraction network, the amount of calculation is reduced while ensuring the ability to extract defects in the liquid crystal display, thereby improving the detection speed of the model; a cross-layer multi-scale feature fusion network is designed to efficiently fuse the shallow fine-grained information and deep semantic information of the feature extraction network; a refined prediction network is designed, including a feature aggregator and a feature purification module. The feature aggregator fuses multi-level features, extracts more contextual information, and enhances the extraction of fine-grained features; the feature purification module suppresses the conflicting information of feature maps of different scales, reduces semantic differences, and improves the accuracy of the model in detecting defects in liquid crystal displays.

[0005] However, the above technology has at least the following technical problems: the existing technology lacks high-sensitivity recognition of complex image backgrounds and tiny defect features, resulting in inaccurate and comprehensive detection of defects such as dead spots, bright spots and color differences in edge and corner areas; it fails to effectively adapt to different ambient lighting conditions, and changes in ambient lighting can easily affect the accuracy and stability of defect detection; it lacks dynamic parameter adjustment and feature enhancement mechanisms, making it difficult to adaptively process diverse image features and defect types. Summary of the invention

[0006] The present invention provides an ultra-thin narrow-frame touch LCD screen image detection method, which solves the technical problems that the prior art lacks high-sensitivity recognition of complex image backgrounds and tiny defect features, resulting in inaccurate and incomplete detection of defects such as dead spots, bright spots and color differences in edge and corner areas; fails to effectively adapt to different ambient lighting conditions, and changes in ambient lighting easily affect the accuracy and stability of defect detection; lacks dynamic parameter adjustment and feature enhancement mechanisms, and is difficult to adaptively process diverse image features and defect types. The method achieves high-precision detection of tiny defects on ultra-thin narrow-frame touch LCD screens under various lighting conditions and optimizes user visual experience.

[0007] The present invention provides an ultra-thin narrow-frame touch-controlled liquid crystal display screen image detection method, which specifically includes the following technical solutions:

[0008] A method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen comprises the following steps:

[0009] S1. By combining high-order dynamic recursive learning with hybrid feature adaptive enhancement technology, a hybrid progressive dynamic learning network is constructed; the feature map obtained by the high-order dynamic recursive learning model is enhanced using an adaptive feature enhancement strategy;

[0010] S2. Introduce an adaptive illumination adjustment algorithm to dynamically adjust the brightness and contrast of the image; further adopt a dynamic content adjustment algorithm based on image processing to identify and process defective areas in the image.

[0011] Preferably, the S1 specifically includes:

[0012] In the implementation of the high-order dynamic recursive learning model, a dynamic recursive formula is defined based on a feature generation function to iteratively update the feature map; the feature generation function depends on the feature map of the current iteration and the dynamic parameters.

[0013] Preferably, the S1 further includes:

[0014] In the process of implementing the adaptive feature enhancement strategy, an adaptive weight coefficient is introduced to dynamically adjust the contribution of different feature sources according to the importance of the feature.

[0015] Preferably, the S2 specifically includes:

[0016] An adaptive illumination adjustment algorithm is introduced to dynamically adjust the brightness and contrast of the image by analyzing the brightness level and local contrast of the image on the ultra-thin narrow-frame touch LCD display.

[0017] Preferably, the S2 further includes:

[0018] In the process of implementing the adaptive illumination adjustment algorithm, a method based on image color space transformation is introduced to detect the ambient light intensity by calculating the average brightness of the image.

[0019] Preferably, the S2 further includes:

[0020] Based on the average brightness and local contrast information of the image, an illumination adjustment factor is designed; and based on the illumination adjustment factor, the brightness and contrast of the image are adjusted.

[0021] Preferably, the S2 further includes:

[0022] In the process of processing the defective areas in the identified images, content reconstruction and dynamic masking methods are introduced.

[0023] Preferably, the S2 further includes:

[0024] In the process of performing content reorganization and dynamic masking, the adjusted image content is obtained by adjusting the masking effect and the influence degree of the mask.

[0025] The beneficial effects of the technical solution of the present invention are:

[0026] 1. By combining high-order dynamic recursive learning with hybrid feature adaptive enhancement technology, the hybrid progressive dynamic learning network can effectively identify tiny defects in the edges and corners of the display, such as dead spots, bright spots and color aberration. Through training with a large number of sample images, it not only improves the recognition accuracy, but also strengthens the generalization ability on different types of LCD screens and complex backgrounds.

[0027] 2. The adaptive illumination adjustment algorithm can dynamically adjust the brightness and contrast of the image according to the different ambient illumination conditions and background scenes simulated in the laboratory, ensuring that the accuracy and stability of defect detection are not affected by changes in ambient illumination, and having higher reliability and robustness in practical applications;

[0028] 3. Through the dynamic content adjustment algorithm based on image processing, the defective areas in the image are accurately identified, and content reorganization and dynamic masking are performed to visually reduce the impact of defects. This not only minimizes the interference of the detection process on normal use and display effects, but also significantly optimizes the user's visual experience. In the ultra-thin narrow frame design, the aesthetics and functionality of the LCD display are maintained;

[0029] 4. Through dynamic recursive formulas and adaptive feature enhancement strategies, the iterative update and optimization of feature maps are realized, and the contribution of different feature sources can be dynamically adjusted according to the importance of the features, which greatly improves the detection sensitivity of tiny defects and the adaptability to complex image backgrounds. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flow chart of a method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0033] The following is a detailed description of a method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen provided by the present invention in conjunction with the accompanying drawings.

[0034] See attached Figure 1 , which shows a flow chart of a method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen provided by an embodiment of the present invention, comprising the following steps:

[0035] S1. By combining high-order dynamic recursive learning with hybrid feature adaptive enhancement technology, a hybrid progressive dynamic learning network is constructed; the feature map obtained by the high-order dynamic recursive learning model is enhanced using an adaptive feature enhancement strategy;

[0036] In the lab, by simulating lighting conditions and background scenes in different environments, various types of LCD screens are photographed or scanned to collect image data containing various known defects. At the same time, image data can also be collected from LCD screen devices in actual use.

[0037] Aiming at the problem of accurate detection of tiny defects or inconsistencies on ultra-thin narrow-frame touch LCD screens, especially in environments with extremely narrow frames, a method that comprehensively considers image complexity and feature dynamics is adopted. By combining high-order dynamic recursive learning with hybrid feature adaptive enhancement technology, a hybrid progressive dynamic learning network is designed to identify tiny defects such as dead spots, bright spots and color aberrations in the edges and corners of ultra-thin narrow-frame touch LCD screens. Through training on a large number of sample images, not only the recognition accuracy and generalization ability are improved, but also the technical challenges of accurate detection are effectively addressed.

[0038] In the implementation of the high-order dynamic recursive learning model, the depth and width of feature extraction are gradually optimized through a layer-by-layer iterative feedback mechanism to better adapt to extremely small defect features and complex image backgrounds. The dynamic recursive formula gradually refines the feature representation through multiple iterations to capture finer image details. The iterative update of the feature map is achieved by setting a feature generation function, which depends on the feature map of the current iteration and a set of dynamic parameters; the dynamic parameters are continuously updated during the iteration process to adapt to the changes in the features. The dynamic recursive formula is defined as follows:

[0039] F t+1 =F t +P(F t ,Θ t )

[0040] Among them, F t+1 is the feature map of the t+1th iteration, F t is the feature map of the tth iteration, P is the feature generation function, which depends on F t and the dynamic parameter set Θ t , including parameters such as weights and biases. The feature generation function uses a composite nonlinear transformation, expressed as:

[0041]

[0042] Where σ is the sigmoid activation function; θ n is the weight parameter of the nth layer, φ n and ψ n Represent the n-th order nonlinear transformation function and adaptive adjustment function, respectively. Represents a feature fusion operation, such as element addition or concatenation. The nth-order nonlinear transformation function φ n and the nth-order adaptive adjustment function ψ n Defined by the following formula:

[0043] φ n (F t )=tanh(W n ·F t +b n )

[0044] ψ n (F t )=ReLU(α n ·(W′ ′ ·F t +b′ n )+β n )

[0045] Among them, W n , W′ n is the weight matrix, bn 、b ′ n is the bias term, α n and β n is the adjustment coefficient, ReLU and tanh are the activation functions.

[0046] In order to achieve effective updating of feature maps, by mixing features of different scales and dimensions and using adaptive feature enhancement strategies, features that are beneficial to defect detection are extracted and enhanced, while suppressing noise and irrelevant information. The introduction of adaptive weight coefficients dynamically adjusts the contribution of different feature sources according to the importance of the features, thereby achieving high adaptability to complex image backgrounds and high sensitivity to tiny defects. The implementation formula of the adaptive feature enhancement strategy is:

[0047]

[0048] F fusion =ω⊙F T ⊙σ(Conv depth (F T ))+(1-ω)

[0049] ⊙ReLU(Conv depth (FFT(F T )))

[0050] Among them, F enhanced is the enhanced feature map; F fusion is the fusion feature map; F T is the feature map finally output by the high-order dynamic recursive learning model; E(F fusion ) is F fusion The average value of D 2 (F fusion ) is F fusion The variance of the feature; δ is the adjustment coefficient used to control the strength of the enhancement effect; ∈ is a constant used to ensure that the denominator is not zero; ω is the adaptive weight coefficient, which is dynamically adjusted according to the importance of the feature; Conv depth is a convolution operation used to extract the importance of spatial features; FFT is a fast Fourier transform function; ⊙ is an element-by-element multiplication. Adaptive feature enhancement strategies are used to optimize feature expression, ensuring that useful features can be effectively extracted and enhanced under different image and lighting conditions, further improving the accuracy and robustness of defect detection.

[0051] S2. Introduce an adaptive illumination adjustment algorithm to dynamically adjust the brightness and contrast of the image; further adopt a dynamic content adjustment algorithm based on image processing to identify and process defective areas in the image.

[0052] In order to meet the detection requirements under different ambient lighting conditions, an adaptive lighting adjustment algorithm is introduced. By analyzing the overall brightness level and local contrast of the image of the ultra-thin narrow-frame touch LCD display, the brightness and contrast of the image are dynamically adjusted to ensure that the accuracy and stability of defect detection are not affected by changes in ambient lighting.

[0053] The detection of ambient light intensity adopts a method based on image color space transformation, which calculates the average brightness of the entire image and converts the image from its original color space (RGB color space) to a color space more suitable for brightness analysis, so as to accurately estimate the ambient light intensity and adjust the image brightness and contrast accordingly. The specific formula is:

[0054]

[0055] Among them, L avg is the average brightness of the image, M represents the number of rows of the position coordinates, N represents the number of columns of the position coordinates, R(i,j), G(i,j), B(i,j) are the intensity values ​​of the red, green, and blue channels of the image at position (i,j), respectively, ω R ,ω G ,ω B is the conversion coefficient from RGB color space to grayscale image. The above formula accurately reflects the overall brightness of the image, while taking into account the difference in sensitivity of the human eye to different color brightness.

[0056] In order to adjust the image to adapt to different lighting conditions, a calculation formula for the lighting adjustment factor is designed, which not only considers the average brightness of the image, but also incorporates the local contrast information. The specific formula is:

[0057]

[0058] Where A is the illumination adjustment factor, which is used for subsequent image brightness and contrast adjustment; σ L is the standard deviation of local brightness, representing the degree of change in local brightness of the image; σ G is the standard deviation of global brightness; is a constant used to prevent division by zero; λ is a balancing factor used to adjust the effect of local contrast on the illumination adjustment factor; and Used to control the adjustment range based on average brightness. The light adjustment factor keeps the image visually similar under different lighting conditions, improving the accuracy of defect detection.

[0059] Based on the lighting adjustment factor, the brightness and contrast of the image are adjusted as follows:

[0060]

[0061] Among them, I adj (i, j) is the brightness value of the adjusted image at position (i, j); γ is the contrast adjustment parameter; I(i, j) represents the pixel value of the image at position (i, j); max(I) is the maximum value of the image pixel; min(I) is the minimum value of the image pixel; It is an exponential parameter used to further adjust the contrast. By adjusting the brightness and contrast of the image, it ensures that the image under different ambient lighting conditions maintains a consistent visual effect, providing optimized image input for subsequent defect recognition.

[0062] In terms of defect recognition and content adjustment, a dynamic content adjustment algorithm based on image processing is used. First, the defect area in the image is identified, and then the defect area is specially processed to reduce the impact on the user's visual experience. The identification of the defect area uses gradient information, taking into account the detail changes of the enhanced feature map. The specific formula is:

[0063]

[0064] Wherein, D(i,j) represents the binary indicator function of the defect area at the position (i,j) of the image. If the position (i,j) is identified as a defect, D(i,j) = 1; otherwise, D(i,j) = 0; and They represent the enhanced feature maps F enhanced The gradient in the x and y directions, the gradient calculation reflects the degree of change of image details and is used to accurately identify defective areas; ζ is the threshold set according to expert experience, which is used to determine whether an area is considered a defect due to its gradient size.

[0065] For the identified defective areas, content reorganization and dynamic masking methods are used to dynamically adjust the masking effect according to the size and shape of the defective area, and the degree of influence of the mask is adjusted through an exponential decay function, making the content adjustment both natural and effective. While retaining the image details, the impact of the defects on the visual effects is minimized, optimizing the user experience of the display. The specific process is as follows:

[0066]

[0067]

[0068] Where Q(i,j) is the value of the dynamic mask at the image position (i,j); ξ is the intensity value of the color or pattern used to generate the mask; C adj (i, j) is the image content at the final adjusted image position (i, j); η is the adjustment factor of the dynamic mask, which controls the degree of fusion between the mask and the original image; It is a parameter used to adjust the influence depth of the dynamic mask so that the mask effect can be more naturally integrated into the original image; max(D) represents the maximum value of the binary indicator function of the defect area.

[0069] Adjust the layout and dynamic elements of the displayed content in real time, such as through intelligent cropping, content reorganization or dynamic masking, to visually reduce the impact of defects. This not only solves the problem of minimal interference with normal use and display effects when detecting defects, but also significantly optimizes the user's visual experience.

[0070] In summary, a method for detecting an image of an ultra-thin narrow-frame touch-sensitive LCD display screen is completed.

[0071] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0072] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0073] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for detecting an image on an ultra-thin narrow-frame touch-screen LCD display, characterized in that: The following steps are involved: S1. By combining high-order dynamic recursive learning with hybrid feature adaptive enhancement technology, a hybrid recursive dynamic learning network is constructed. In the implementation of the high-order dynamic recursive learning model, a dynamic recursive formula is defined based on the feature generation function to iteratively update the feature map; the feature generation function depends on the feature map and dynamic parameters of the current iteration; the dynamic recursive formula is defined as follows: F t+1 =F t +P(F t ,Θ t ) Among them, F t+1 is the feature map of the t+1th iteration, F t is the feature map of the tth iteration, P is the feature generation function, which depends on F t and the dynamic parameter set Θ t , including parameters such as weights and biases; the feature generation function uses a composite nonlinear transformation, expressed as: Where σ is the sigmoid activation function; θ n is the weight parameter of the nth layer, φ n and ψ n Represent the nonlinear transformation function and adaptive adjustment function of the nth order, Represents feature fusion operation; nth order nonlinear transformation function φ n and the nth-order adaptive adjustment function ψ n Defined by the following formula: φ n (F t )=tanh(W n ·F t +b n ) ψ n (F t )=ReLU(a n ·(W n ′·F t +b′ n )+b n ) Among them, W n , W′ n is the weight matrix; b n , b′ n is the bias term; α n and β n is the adjustment coefficient; ReLU and tanh are activation functions; For the feature map obtained by the high-order dynamic recursive learning model, the adaptive feature enhancement strategy is used for enhancement processing; in the implementation process of the adaptive feature enhancement strategy, the adaptive weight coefficient is introduced to dynamically adjust the contribution of different feature sources according to the importance of the feature; the implementation formula of the adaptive feature enhancement strategy is: F fusion =ω⊙F T ⊙σ(Conv depth (F T ))+(1-ω) ⊙ReLU(Conv depth (FFT(F T ))) Among them, F enhanced is the enhanced feature map; F fusion is the fusion feature map; F T It is the feature map finally output by the high-order dynamic recursive learning model; E(F fusion ) is F fusion The average value of D 2 (F fusion ) is F fusion The variance of Conv is: δ is the adjustment coefficient; ∈ is a constant; ω is the adaptive weight coefficient; dept is a convolution operation; FFT is a fast Fourier transform function; ⊙ is an element-by-element multiplication; S2. Introduce an adaptive lighting adjustment algorithm to dynamically adjust the brightness and contrast of the image; in the process of implementing the adaptive lighting adjustment algorithm, introduce a method based on image color space transformation to detect the ambient light intensity by calculating the average brightness of the image; Based on the average brightness and local contrast information of the image, the illumination adjustment factor is designed. The specific formula is: Where A is the illumination adjustment factor; σ L is the standard deviation of local brightness; σ G is the standard deviation of global brightness; is a constant; λ is a balancing factor; and It is used to control the adjustment amplitude based on average brightness; And based on the illumination adjustment factor, adjust the brightness and contrast of the image, which is specifically implemented as follows: Among them, I adj (i, j) is the brightness value of the adjusted image at position (i, j); γ is the contrast adjustment parameter; I(i, j) represents the pixel value of the image at position (i, j); L avg is the average brightness of the image; max(I) is the maximum value of the image pixel; min(I) is the minimum value of the image pixel; is an exponential parameter used to further adjust the contrast; A dynamic content adjustment algorithm based on image processing is further adopted to identify and process defective areas in the image.

2. The method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen according to claim 1, characterized in that: The S2 specifically includes: An adaptive illumination adjustment algorithm is introduced to dynamically adjust the brightness and contrast of the image by analyzing the brightness level and local contrast of the image on the ultra-thin narrow-frame touch LCD display.

3. The method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen according to claim 1, characterized in that: Said S2 further comprises: In the process of processing the defective areas in the identified images, content reconstruction and dynamic masking methods are introduced.

4. The method for detecting an image on an ultra-thin narrow-frame touch-controlled liquid crystal display screen according to claim 3, characterized in that: Said S2 further comprises: In the process of performing content reorganization and dynamic masking, the adjusted image content is obtained by adjusting the masking effect and the influence degree of the mask.

Images