A method for evaluating the surface quality of a brush-polished screen glass substrate

Abstract

The present invention relates to the technical field of glass substrate quality analysis, and in particular to a method for evaluating the surface quality of a brush-polished screen glass substrate. The method includes: scanning the surface of the screen glass substrate to obtain raw image data, and performing intelligent preprocessing on the raw image data to obtain preprocessed image data; based on the preprocessed image data, introducing multi-dimensional feature coupling analysis and dynamic weighting algorithm to extract and optimize features to generate comprehensive feature data; based on the comprehensive feature data, introducing multi-layer feature coupling weighting and adaptive multi-dimensional evaluation algorithm to perform intelligent analysis and quality evaluation classification to obtain the quality category of the substrate surface. The method solves the problems of inaccurate processing of substrate surface data, insufficient and inaccurate analysis of substrate characteristics in the quality evaluation of the brush-polished screen glass substrate surface.

Description

Technical Field

[0001] The invention relates to the technical field of glass substrate quality analysis, and in particular to a method for evaluating the surface quality of a brush-polished screen glass substrate. Background Art

[0002] In the manufacturing of modern electronic products, the surface quality of screen glass substrates directly affects the display effect and user experience, especially in application scenarios with high requirements for touch and optical performance, such as smartphones, tablets and monitors. Screen glass substrates usually need to go through multiple polishing processes to ensure surface flatness, smoothness and minimize particle residue. However, subtle surface defects, including scratches, residual particles, uneven polishing, etc., will inevitably appear during the polishing process. These defects may affect the appearance and performance of the final product, so strict quality control is required.

[0003] Existing quality inspection methods usually use optical imaging technology combined with manual inspection or simple image processing algorithms. However, these traditional methods have obvious limitations in high-precision inspection and automation, especially when dealing with details such as fine particle residues and microscopic scratches, it is difficult to achieve ideal results. Traditional inspection methods rely on a single image feature, such as a simple threshold segmentation method or edge detection algorithm based on grayscale values, which cannot fully capture complex surface features, resulting in low inspection accuracy and difficulty in automating feature extraction and comprehensive evaluation. At the same time, in the surface quality assessment of brush-polished screen glass substrates, there are problems such as inaccurate processing of substrate surface data, insufficient analysis of substrate characteristics, and inaccurate analysis. Summary of the invention

[0004] The invention provides a method for evaluating the surface quality of a brush-polished screen glass substrate, so as to solve the problems of inaccurate substrate surface data processing, insufficient and inaccurate substrate characteristic analysis in the brush-polished screen glass substrate surface quality evaluation.

[0005] A method for evaluating the surface quality of a brush-polished screen glass substrate comprises the following steps:

[0006] S1. Scan the surface of the screen glass substrate to obtain the original image data, and perform intelligent preprocessing on the original image data. In the intelligent preprocessing process, the original image data is denoised to obtain the image data after the noise is removed; introduce a multi-order composite grayscale transformation, perform grayscale processing on the image data after the noise is removed, and obtain the image data after the multi-order composite grayscale transformation; the formula of the multi-order composite grayscale transformation is as follows:

[0007]

[0008] in, is the image after noise removal at coordinate position The gray value at ; It is the image after multi-level composite grayscale transformation at coordinate position The gray value at ; is the grayscale adjustment factor; is the nonlinear frequency control parameter; is the logarithmic transformation gain parameter; is the global grayscale mean of the image after noise removal; is the factor of periodic change;

[0009] A random noise model is introduced to optimize the image data after multi-order composite grayscale transformation to obtain optimized image data; edge enhancement processing is performed on the optimized image data to obtain pre-processed image data;

[0010] Based on the preprocessed image data, multi-dimensional feature coupling analysis and dynamic weighting algorithm are introduced to extract and optimize features to generate comprehensive feature data;

[0011] S2. Based on the comprehensive feature data, a multi-layer feature coupling weighting and adaptive multi-dimensional evaluation algorithm are introduced to perform intelligent analysis and quality assessment classification to obtain the quality category of the substrate surface.

[0012] Preferably, the S1 specifically includes:

[0013] In the process of implementing the multi-dimensional feature coupling analysis and dynamic weighting algorithm, the roughness features, particle residual density, smoothness features and uniformity features are extracted; and the roughness features, particle residual density, smoothness features and uniformity features are multi-dimensionally coupled and dynamically weighted to generate a comprehensive quality assessment value.

[0014] Preferably, the S1 specifically includes:

[0015] The calculation formula of the comprehensive quality assessment value is as follows:

[0016]

[0017] in, is the comprehensive quality assessment value; , , is the dynamic weight coefficient; Represents roughness characteristics; is the particle residual density; Indicates smoothness characteristics; It is a uniformity feature;

[0018] Based on roughness characteristics, particle residual density, smoothness characteristics, uniformity characteristics and comprehensive quality evaluation values, comprehensive characteristic data are constructed.

[0019] Preferably, the S2 specifically includes:

[0020] The implementation process of multi-layer feature coupling weighting and adaptive multi-dimensional evaluation algorithm is as follows:

[0021] The first step is to construct a feature matrix and perform multi-layer feature coupling weighting processing to obtain coupling features;

[0022] The second step is to construct an adaptive weighted feature matrix based on the coupling features;

[0023] The third step is to perform adaptive multi-dimensional evaluation processing based on the adaptive weighted feature matrix to generate a multi-dimensional quality evaluation value of the substrate surface;

[0024] The fourth step is to classify and judge the quality based on the multi-dimensional quality evaluation value of the substrate surface.

[0025] Preferably, the S2 specifically includes:

[0026] Based on the feature matrix, the feature coupling function is introduced , the coupling between features is performed by multi-layer coupling and weighted method to generate coupling features; the calculation formula for any element in the coupling feature is as follows:

[0027]

[0028] in, It is The coupling components of the features; and They represent the first and The characteristic components include roughness characteristics , particle residual density , Smoothness characteristics , uniformity characteristics and comprehensive quality assessment value ; is the coupling weight coefficient between features; is the coupling adjustment factor.

[0029] Preferably, the S2 specifically includes:

[0030] Design a multidimensional evaluation function based on an adaptive weighted feature matrix , generate the multi-dimensional quality evaluation value of the substrate surface; the specific calculation formula is as follows:

[0031]

[0032] in, is the multidimensional quality evaluation value of the substrate surface; is the adaptive weighted feature matrix; is the exponential decay coefficient; is the adaptive weighted feature matrix Line Elements of a column; is an exponential decay term; is the normalization term.

[0033] Preferably, the S2 specifically includes:

[0034] Based on the multi-dimensional quality evaluation value of the substrate surface, an adaptive classification judgment function is used Automatic classification based on quality judgment threshold; adaptive classification judgment function The definition is as follows:

[0035]

[0036] in, and is the quality judgment threshold.

[0037] The beneficial effects of the technical solution of the present invention are:

[0038] 1. The surface image data of the screen glass substrate is collected through an optical detection device, and intelligent preprocessing is performed, including Gaussian filtering and median filtering denoising, multi-order composite grayscale transformation and random noise enhancement. Multi-order composite grayscale transformation can enhance the contrast of the image, making the details of dark and bright areas more prominent, ensuring the accuracy of feature extraction. At the same time, noise enhancement processing simulates the microscopic distribution of particles on the surface of the substrate, further enhancing the sensitivity to particle details, and laying the foundation for the subsequent accurate extraction of features. Through a series of efficient image processing methods, accurate capture of substrate surface details and data optimization are achieved, effectively improving detection accuracy.

[0039] 2. Through the multi-dimensional coupling of the gray value, position angle, gray gradient and gray curvature of the pixel points of the preprocessed image data, the roughness, residual particle density, smoothness and uniformity are extracted. The calculation formula of each feature is combined with complex nonlinear functions, thus establishing a multi-level coupling relationship between different features.

[0040] 3. Introduce multi-layer feature coupling weighting and adaptive multi-dimensional evaluation algorithm for intelligent analysis and quality assessment classification. Generate an adaptive weighted feature matrix to enhance the adaptability and adjustability of the coupling feature. The adaptive weighted feature matrix is ​​constructed through the interactive relationship between the coupling component of each feature and the feature component in the feature matrix, realizing the adaptive adjustment of multi-level features, avoiding the phenomenon of excessive or too small gradients in the data processing process, so that each feature component in the adaptive weighted feature matrix maintains a unique and relatively balanced expression effect. Through the adaptive weighted control between features, it is ensured that the weight of the feature can be automatically adjusted under different substrate surface conditions, which significantly improves the robustness and adaptability of the evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The present invention provides a flow chart of a method for evaluating the surface quality of a brush-polished screen glass substrate. DETAILED DESCRIPTION

[0042] 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.

[0043] 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.

[0044] The specific scheme of the surface quality assessment method of a brush-polished screen glass substrate provided by the present invention is described in detail below with reference to the accompanying drawings.

[0045] See attached Figure 1 , which shows a flow chart of a method for evaluating the surface quality of a brush-polished screen glass substrate provided by an embodiment of the present invention, the method comprising the following steps:

[0046] S1. Scan the surface of the screen glass substrate to obtain raw image data, and perform intelligent preprocessing on the raw image data to obtain preprocessed image data; based on the preprocessed image data, introduce multi-dimensional feature coupling analysis and dynamic weighting algorithm to extract and optimize features to generate comprehensive feature data.

[0047] After polishing, a high-resolution optical detection device (such as a CCD camera or a CMOS camera) is used to comprehensively scan the surface of the screen glass substrate to obtain original image data.

[0048] The original image data is preprocessed intelligently. The specific implementation process is as follows:

[0049] First, the existing filtering algorithm is used to remove noise, such as using a convolution operation with a Gaussian kernel to smooth the original image data to implement Gaussian filtering to remove high-frequency noise; a specified window is used to sort each pixel and take the median, retaining the edge features to implement median filtering, remove speckle noise, and obtain the image data after noise removal.

[0050] Furthermore, a multi-order composite grayscale transformation is introduced to grayscale the image data after noise removal; through the multi-order composite grayscale transformation, the contrast of different grayscale intervals in the image after noise removal is enhanced, which is convenient for the subsequent accurate extraction of features. The goal of the multi-order composite grayscale transformation is to remap the grayscale value of each pixel of the image after noise removal using a nonlinear relationship, so that the image after noise removal can show significant details in both darker and brighter areas. The formula of the multi-order composite grayscale transformation is as follows:

[0051]

[0052] in, is the image after noise removal at coordinate position The gray value at ; It is the image after multi-level composite grayscale transformation at coordinate position The gray value at ; is the overall grayscale adjustment coefficient, which is used to control the brightness of the output image; It is a nonlinear frequency control parameter, which is used to adjust the frequency of grayscale changes; It is the logarithmic transformation gain parameter, which is used to control the brightness range expansion; It is the global grayscale mean of the image after noise removal, which is used for local brightness balance. It is a periodically changing factor, which is used to make the grayscale value of the image after noise removal have a larger contrast difference in different brightness ranges (highlight area and dark area), thereby enhancing the details of the low contrast area.

[0053] After the above multi-order composite grayscale transformation, in order to further enhance the surface details of the substrate, a random noise model is introduced to optimize the image data after the multi-order composite grayscale transformation; by introducing noise with specific frequency and spatial characteristics into the image after the multi-order composite grayscale transformation, the microscopic distribution of particles on the substrate surface is simulated, making the subsequent feature extraction more sensitive to particles. In the random noise model, Gaussian distribution is used to control the spatial range of noise, and sine function is used to control the frequency change of noise. The specific formula of the random noise model is:

[0054]

[0055] in, It is the image after multi-level composite grayscale transformation at coordinate position The noise value at is the noise amplitude, which is used to control the brightness and contrast of the noise in the image after multi-level composite grayscale transformation; is the center coordinate of the noise, which specifies the starting position of the noise distribution. By changing the center position, the particle distribution in different areas can be simulated; is the scale parameter of the Gaussian distribution, which is used to control the range of noise diffusion; is the noise frequency, used to adjust the noise periodicity; is the spatial diffusion term controlled by the Gaussian distribution; is the periodic variation of noise, which is used to control the spatial distribution frequency of noise. The above random noise model combines the characteristics of Gaussian distribution and sine wave, and can simulate a random noise distribution with multi-scale characteristics. The optimized image data is obtained .

[0056] Furthermore, the optimized image data is subjected to edge enhancement processing using existing technologies. For example, the Laplace operator is used to detect high-frequency parts in the optimized image, such as small scratches and residual particles; the Sobel operator is used to further highlight edge features by calculating the gradient size and direction, which helps to identify particles and scratches. The image data after edge enhancement processing, i.e. the preprocessed image data, is obtained. .

[0057] Furthermore, based on the preprocessed image data , introduces multi-dimensional feature coupling analysis and dynamic weighting algorithm to realize feature extraction and optimization. Multi-dimensional feature coupling analysis and dynamic weighting algorithm extracts features from multiple dimensions such as roughness, residual particle density, smoothness and uniformity, and generates comprehensive feature data after multi-dimensional feature coupling and dynamic weighting of features. The specific implementation process is as follows:

[0058] First, multi-level features are extracted, including roughness, residual particle density, smoothness, and uniformity. In each step of feature extraction, indicators of different dimensions such as differential operation, angle distribution, grayscale mean and variance are combined to form a multi-level surface feature expression, and finally a comprehensive quality assessment value is generated through coupling weighting.

[0059] In order to extract the roughness characteristics of the substrate surface, the grayscale value of the pixel point of the preprocessed image is differentiated to extract the microscopic roughness characteristics. In the pixel space of the substrate image, the grayscale value of any pixel point of the preprocessed image is set to , calculate the gradient size of the pixel to characterize the roughness of the pixel. The specific formula is:

[0060]

[0061] in, It represents the roughness characteristics and is a comprehensive numerical expression of the surface roughness of the substrate; is the total number of pixels in the preprocessed image; is the first The gray value of each pixel is used to reflect the brightness information; is the first The angle parameter of pixels is used to adjust and The influence of directional gray gradient; The first image after preprocessing Pixels in The gray gradient in the direction, indicating Grayscale change rate in direction; The first image after preprocessing Pixels in The gray gradient in the direction, indicating Grayscale change rate in direction; and They represent the first Pixels in and Directional gradient weighting is used to adjust the impact of grayscale gradients in different directions on roughness; ,in The first image after preprocessing The coordinate position of the pixel point; The term is used to amplify the effect of gray value changes on roughness; The influence weight of high grayscale areas is enhanced by introducing exponential terms. The above roughness characteristic formula combines the position angle, grayscale gradient, logarithm of grayscale value and exponential terms to achieve a multi-dimensional expression of roughness characteristics, so that the roughness of the substrate surface can be accurately characterized.

[0062] Furthermore, tiny particles often remain on the substrate surface after polishing, and the accurate calculation of the particle residue density is crucial for the evaluation of the substrate surface quality. By identifying the area and position angle of each particle area and combining specific logarithmic and trigonometric functions, the calculation formula for the particle residue density is defined:

[0063]

[0064] in, is the particle residual density, which is a comprehensive numerical representation of the particle distribution per unit area on the substrate surface; Indicates the total area of ​​the detection area; is the total number of particles detected; is the sine value of the particle position angle, which is used to weight the particle residual density. ; For the The coordinate position of each particle; It is The area of ​​each particle reflects the size of the particle; It is a logarithmic term based on the particle area, reflecting the particle distribution and density characteristics. The above formula accurately reflects the density and residual characteristics of particle distribution by combining particle area, position angle, logarithm and trigonometric function terms in the calculation of particle residual density, providing numerical support for the accurate evaluation of particle residual density.

[0065] Furthermore, the smoothness feature is extracted based on the relationship between the grayscale change curvature (i.e., the second-order derivative of the grayscale) and the grayscale change rate. The second-order derivative of the grayscale reflects the microscopic change of the substrate surface smoothness, which is combined with the previously extracted particle residual density. Coupled to form a comprehensive evaluation of the substrate surface smoothness. The smoothness formula is as follows:

[0066]

[0067] in, It represents the smoothness characteristic and is a numerical representation of the smoothness of the substrate surface; The first image after preprocessing Pixels in The second-order derivative of the grayscale in the direction, expressed as The curvature of grayscale change in the direction; is the first Pixels in The second-order derivative of the grayscale in the direction, expressed as The curvature of the grayscale change in the direction; Combined with and The negative exponential term of the first-order derivative of the directional grayscale weakens the weight of the high gradient area and highlights the contribution of the low gradient (smooth area) to the smoothness; The first image after preprocessing Pixels in The gray gradient in the direction, indicating Grayscale change rate in direction; The first image after preprocessing Pixels in The gray gradient in the direction, indicating Grayscale change rate in direction; is the first The gray value of each pixel; Based on the square of the gray value and the residual density of particles The logarithmic term of is combined with the above-mentioned residual particle density characteristics to further amplify the influence of high grayscale areas on smoothness. The above smoothness formula combines the grayscale gradient, the curvature of grayscale change and the residual particle density characteristics to form a complex expression of smoothness, ensuring a multi-level evaluation of the smoothness of the substrate surface.

[0068] Furthermore, the uniformity feature is used to measure the uniformity of the grayscale distribution on the substrate surface. Coupling to obtain uniformity index , the expression is as follows:

[0069]

[0070] in, It is a uniformity feature, indicating the uniformity of the surface grayscale distribution; is the number of blocks within the substrate detection area; It is The modulus of the grayscale gradient of the pixels in the block area is and The grayscale gradients in different directions are merged to fully reflect the grayscale changes; It is Gray value of the pixel points in the block area; It is a power operation on the gradient modulus value to adjust the sensitivity of smoothness; is the power exponent, which controls the sensitivity of the response to smoothness; It is The grayscale mean of the pixels in the block area; It is The pixel grayscale variance within the block area; It is a trigonometric function combining smoothness and grayscale variance to enhance the sensitivity to uniformity features. The above uniformity formula constructs a complex evaluation of substrate surface uniformity by coupling grayscale distribution, smoothness and block variance, ensuring multi-dimensional feedback on substrate surface quality.

[0071] Furthermore, after completing the feature extraction of each dimension, the roughness feature, particle residual density, smoothness feature and uniformity feature are multi-dimensionally coupled and dynamically weighted to generate a comprehensive quality assessment value. By combining the structure of nonlinear dynamic weighting, the following formula is formed:

[0072]

[0073] in, It is a comprehensive quality assessment value used to characterize the surface quality of the substrate; , , is a dynamic weight coefficient, determined according to the specific scenario; It is the negative exponential term of the smoothness feature, which is used to control the influence of the roughness feature on the comprehensive quality; It is the square root of the uniformity characteristic and is used to dynamically adjust the comprehensive quality evaluation value. The above formula realizes a comprehensive evaluation of the substrate surface quality through nonlinear weighting and dynamic adjustment under multi-dimensional coupling.

[0074] After the above processing, comprehensive characteristic data are obtained, including roughness characteristics, particle residual density, smoothness characteristics, uniformity characteristics and comprehensive quality evaluation values.

[0075] S2. Based on the comprehensive feature data, a multi-layer feature coupling weighting and adaptive multi-dimensional evaluation algorithm are introduced to perform intelligent analysis and quality assessment classification to obtain the quality category of the substrate surface.

[0076] Based on comprehensive feature data, a multi-layer feature coupling weighted and adaptive multi-dimensional evaluation algorithm is introduced for intelligent analysis and quality assessment classification. The multi-layer feature coupling weighted and adaptive multi-dimensional evaluation algorithm is not only based on complex feature coupling and weighted calculation, but also forms a dynamic evaluation structure through layer-by-layer adaptive weight adjustment to achieve accurate classification of substrate surface quality. The specific implementation process is as follows:

[0077] The first step is to construct a feature matrix and perform multi-layer feature coupling weighting processing;

[0078] In order to quantify the substrate surface features from different angles, a feature matrix is ​​constructed The feature matrix contains multiple key substrate surface feature data, namely comprehensive feature data, including roughness feature, particle residue density, smoothness feature, uniformity feature and comprehensive quality evaluation value. The comprehensive feature data reflects the quality parameters of different aspects of the substrate surface and constitutes the basic data of the evaluation process.

[0079] Feature Matrix Defined as:

[0080]

[0081] On this basis, the characteristic coupling function is introduced To describe the weighted coupling relationship between the features, the coupled features are represented by multi-layer coupling and weighted weighting, thereby generating a comprehensive coupling feature. In order to enhance the complexity of the coupling feature, sine, cosine and exponential operations are introduced to generate any element in the following coupling feature, namely the coupling component:

[0082]

[0083] in, It is The coupling component of the first feature, The coupling effect between a feature and other features; and They represent the first and The characteristic components include roughness characteristics , particle residual density , Smoothness characteristics , uniformity characteristics and comprehensive quality assessment value ; is the coupling weight coefficient between features, which is used to adjust the influence strength of different features to cope with the importance of different features in quality assessment; It is a sine term based on the product of eigencomponents and is used to express the nonlinear relationship between eigencomponents; It is the cosine function of the difference between the feature components. The difference between the feature components is mapped to a value in the range of [-1,1] through the cosine to enhance the sensitivity to the difference. It is the exponential term of the product of characteristic components, which is used to increase the complexity and sensitivity of the coupling relationship; is the coupling adjustment coefficient, and its value is determined through experimental data to adaptively adjust the effect of coupling between features. Each term in the above formula is a complex expression of the coupling relationship between features, ensuring that the value and variation of each feature can fully demonstrate its impact on the surface quality of the substrate through a nonlinear relationship.

[0084] The second step is to construct an adaptive weighted feature matrix based on the coupling features;

[0085] In obtaining the coupled eigenvalues After that, we need to construct an adaptive weighted feature matrix , which is used to generate the final comprehensive eigenvalue. The adaptive weighting mechanism is used to adjust the features at multiple levels, so that the adaptive weighted feature matrix generated in the end can fully reflect the quality characteristics of the substrate surface.

[0086] Adaptive weighted feature matrix The calculation formula for any element of is as follows:

[0087]

[0088] in, is the adaptive weighted feature matrix Line The elements of the column are used as input for subsequent quality assessment processing; Represents the variance between features, which is used to smooth the adaptive weighted feature matrix to avoid excessive feature gradients; is the inverse tangent operation of the characteristic components and the coupling matrix terms, which enhances the nonlinear relationship and ensures that different features still maintain unique numerical distributions after multi-layer coupling. The adaptive weighted feature matrix , which is accomplished through cross-coupling and adaptive weighted calculation between multi-dimensional features, thus forming a multi-layer complex feature expression within the matrix dimension.

[0089] The third step is to perform adaptive multi-dimensional evaluation processing based on the adaptive weighted feature matrix;

[0090] Get the adaptive weighted feature matrix After that, further, based on Designing multidimensional evaluation functions , generate a multi-dimensional quality evaluation value of the substrate surface, and realize a comprehensive judgment on the substrate surface quality.

[0091] The calculation formula of the multi-dimensional quality evaluation value of the substrate surface is as follows:

[0092]

[0093] in, It is the final multi-dimensional quality evaluation value of the substrate surface, which is used to judge the quality grade of the substrate surface; is the exponential decay coefficient, which is used to reduce the influence of some noise features and make the multidimensional evaluation process more robust; is an exponential decay term that helps to attenuate the effects of high noise features; It is a normalization term, which is used to normalize the components in the adaptive weighted feature matrix to ensure the stability of the distribution of multidimensional quality evaluation values; It is a logarithmic function of the normalized term, which is used to increase the sensitivity to the substrate surface. Through the multidimensional evaluation function, each element in the adaptive weighted feature matrix can be processed layer by layer to generate the final multidimensional quality evaluation value.

[0094] Step 4: Classify and judge quality based on multi-dimensional quality evaluation values;

[0095] In order to obtain the multidimensional quality evaluation value Then, the adaptive classification decision function is used Based on quality threshold and Automated classification is performed to determine the exact quality level of the substrate surface.

[0096] The adaptive classification decision function is defined as follows:

[0097]

[0098] in, It is an adaptive classification judgment function, which is used to finally output the quality category of the substrate surface; and It is the threshold for quality judgment, which is set adaptively according to experimental data and is used to determine the boundaries between different categories.

[0099] In summary, a surface quality assessment method for brush-polished screen glass substrates was completed.

[0100] 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.

[0101] 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.

[0102] 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 evaluating the surface quality of a brush-polished screen glass substrate, characterized in that: The following steps are involved: S1. Scan the surface of the screen glass substrate to obtain the original image data, and perform intelligent preprocessing on the original image data. In the intelligent preprocessing process, the original image data is denoised to obtain the image data after the noise is removed; introduce a multi-order composite grayscale transformation, perform grayscale processing on the image data after the noise is removed, and obtain the image data after the multi-order composite grayscale transformation; the formula of the multi-order composite grayscale transformation is as follows: in, is the image after noise removal at coordinate position The gray value at ; It is the image after multi-level composite grayscale transformation at coordinate position The gray value at ; is the grayscale adjustment factor; is the nonlinear frequency control parameter; is the logarithmic transformation gain parameter; is the global grayscale mean of the image after noise removal; is the factor of periodic change; A random noise model is introduced to optimize the image data after multi-order composite grayscale transformation to obtain optimized image data; edge enhancement processing is performed on the optimized image data to obtain pre-processed image data; Based on the preprocessed image data, multi-dimensional feature coupling analysis and dynamic weighting algorithm are introduced to extract and optimize features to generate comprehensive feature data; S2. Based on the comprehensive feature data, a multi-layer feature coupling weighting and adaptive multi-dimensional evaluation algorithm are introduced to perform intelligent analysis and quality assessment classification to obtain the quality category of the substrate surface; The implementation process of multi-layer feature coupling weighting and adaptive multi-dimensional evaluation algorithm is as follows: The first step is to construct a feature matrix and perform multi-layer feature coupling weighting processing to obtain coupling features; The second step is to construct an adaptive weighted feature matrix based on the coupling features; The third step is to perform adaptive multi-dimensional evaluation processing based on the adaptive weighted feature matrix to generate a multi-dimensional quality evaluation value of the substrate surface; The fourth step is to classify and judge the quality based on the multi-dimensional quality evaluation value of the substrate surface; Based on the feature matrix, the feature coupling function is introduced , the coupling between features is performed by multi-layer coupling and weighted method to generate coupling features; the calculation formula for any element in the coupling feature is as follows: in, It is The coupling components of the features; and They represent the first and The characteristic components include roughness characteristics , particle residual density , Smoothness characteristics , uniformity characteristics and comprehensive quality assessment value ; is the coupling weight coefficient between features; is the coupling adjustment coefficient; Design a multidimensional evaluation function based on an adaptive weighted feature matrix , generate the multi-dimensional quality evaluation value of the substrate surface; the specific calculation formula is as follows: in, is the multidimensional quality evaluation value of the substrate surface; is the adaptive weighted feature matrix; is the exponential decay coefficient; is the adaptive weighted feature matrix Line Elements of a column; is an exponential decay term; is the normalization term.

2. The method for evaluating the surface quality of a brush-polished screen glass substrate according to claim 1, characterized in that: The S1 specifically includes: In the process of implementing the multi-dimensional feature coupling analysis and dynamic weighting algorithm, the roughness features, particle residual density, smoothness features and uniformity features are extracted; and the roughness features, particle residual density, smoothness features and uniformity features are multi-dimensionally coupled and dynamically weighted to generate a comprehensive quality assessment value.

3. The method for evaluating the surface quality of a brush-polished screen glass substrate according to claim 2, characterized in that: The S1 specifically includes: The calculation formula of the comprehensive quality assessment value is as follows: in, is the comprehensive quality assessment value; , , is the dynamic weight coefficient; Represents roughness characteristics; is the particle residual density; Indicates smoothness characteristics; It is a uniformity feature; Based on roughness characteristics, particle residual density, smoothness characteristics, uniformity characteristics and comprehensive quality evaluation values, comprehensive characteristic data are constructed.

4. The method for evaluating the surface quality of a brush-polished screen glass substrate according to claim 1, characterized in that: The S2 specifically includes: Based on the multi-dimensional quality evaluation value of the substrate surface, an adaptive classification judgment function is used Automatic classification based on quality judgment threshold; adaptive classification judgment function The definition is as follows: in, and is the quality judgment threshold.

Images