Self-adhesive film adhesive force evaluation method based on pattern recognition

Abstract

The invention relates to the technical field of image processing and pattern recognition, and discloses a self-adhesive film adhesive force evaluation method based on pattern recognition. A contact area height map is obtained through three-dimensional measurement, and zero offset is achieved; projecting the height matrix column mean value to generate a one-dimensional elevation curve; discrete curvatures are extracted through three-point difference, and the total amount is obtained through accumulation; accumulating coefficient squares to obtain total energy by using a self-defined non-orthogonal spectral basis expansion curve; solving the mean value, the standard deviation, the skewness and the kurtosis of the energy sequence; frequency spectrum intensity is generated column by column according to the coefficient and the spectrum basis, a mean value and a standard deviation are taken to identify a salient region, and the average intensity is solved; constructing an adhesive force evaluation function according to the total curvature, the total energy, the skewness and the kurtosis; and forming a normalized adhesion index by combining the evaluation value and the average intensity of the salient region, and grading according to a threshold value.

Description

Technical Field

[0001] This invention relates to the field of image processing and pattern recognition technology, specifically to a method for evaluating the adhesion force of self-adhesive films based on pattern recognition. Background Technology

[0002] Self-adhesive materials are widely used as adhesives, mucosal patches, and controlled-release formulations in moist environments such as the oral cavity, nasal cavity, and digestive tract. Their adhesion to mucosal surfaces directly affects the stability of the drug delivery site, duration of action, and drug bioavailability. To evaluate the performance of self-adhesive materials, current technologies typically employ macroscopic mechanical testing methods such as tensile peel tests, shear adhesion tests, pull-off strength tests, and texture analyzer measurements of adhesive work, or combine these with in vitro mucosal models to record and compare mechanical indicators such as maximum peel force and interfacial separation time.

[0003] On the other hand, with the development of equipment such as confocal microscopy, white light interferometry, and 3D profilometers, some technical solutions have begun to utilize surface morphology measurement results to statistically analyze the roughness parameters of the adhesion interface (such as arithmetic mean roughness, root mean square roughness, peak-valley height, etc.) and correlate these two-dimensional or three-dimensional roughness indices with adhesion performance. However, such methods mostly remain at the level of empirical correlation of a few scalar parameters, without systematically quantifying the distribution characteristics of morphology in the spatial scale and frequency domain, and cannot reflect the differential contribution of local microstructures to adhesion behavior. The aforementioned macroscopic mechanical testing methods usually require the application of external force to the mucosa or mucosa-like carrier, which is somewhat destructive. The test results are easily affected by loading speed, environmental humidity, temperature, and subjective factors of the operator, resulting in poor repeatability and comparability. At the same time, these methods only obtain macroscopic results such as overall adhesion strength or average adhesion work, making it difficult to obtain spatial distribution information of adhesion ability at different locations on the interface, and failing to provide refined morphological guidance for material formulation optimization or surface structure design. Existing analysis methods based on image or topographic data also employ general signal processing or image processing techniques, such as traditional Fourier transform, wavelet decomposition, or simple texture feature extraction on height maps, and then introduce general machine learning or deep learning models for classification or regression. However, such methods often rely on a large number of labeled samples for training, resulting in numerous model parameters and weak interpretability. When the sample distribution changes or extreme smooth or rough edge shapes appear, the output results lack stability. Furthermore, these methods tend to focus on the model's "fitting" ability rather than starting from the combination of physically interpretable quantities such as topographic curvature changes, spectral energy distribution, and local spectral intensity to construct a quantitative index system with clear physical meaning, dimensional consistency, and direct applicability for adhesion level classification.

[0004] Therefore, this study aims to propose a pattern recognition-based method for evaluating the adhesion force of self-adhesive films. First, a surface height map is obtained through 3D scanning, generating a zero-offset height matrix. Then, a one-dimensional elevation curve is formed by projection along the column direction. Curvature changes are extracted using three-point difference analysis, and the elevation curve is then spectrally expanded based on a custom family of non-orthogonal spectral basis functions. Next, statistical analysis is performed on the spectral coefficients to extract higher-order features such as spectral energy, skewness, and kurtosis. Subsequently, a spectral intensity response sequence is mapped, and significant regions are identified. Finally, the total curvature, spectral energy, and spectral distribution features are fused to construct an adhesion force evaluation function and a normalized adhesion index, and adhesion force levels are classified. Summary of the Invention

[0005] This invention provides a pattern recognition-based method for evaluating the adhesion force of self-adhesive films, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for evaluating the adhesion force of self-adhesive films based on pattern recognition, comprising:

[0007] The contact area between the self-adhesive film material and the target adhesion surface is measured in three dimensions to obtain a height map of the contact area, and a zero-offset height matrix is ​​generated based on the height map.

[0008] The zero-offset height matrix is ​​averaged along the column direction to obtain the average height sequence of each column, and a one-dimensional elevation curve is generated.

[0009] On a one-dimensional elevation curve, a three-point difference operation is performed on each intermediate sampling point to obtain a discrete curvature change sequence, and the absolute values ​​of the discrete curvature changes are accumulated to obtain the total curvature change.

[0010] The spectrum of a one-dimensional elevation curve is expanded based on a family of nonorthogonal spectral basis functions. The expansion coefficients of each spectral order are obtained, and the squares of the expansion coefficients of each spectral order are accumulated to obtain the total spectral energy.

[0011] Statistical analysis was performed on the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order to obtain the mean, standard deviation, spectral distribution skewness and kurtosis based on the third and fourth central moments of the spectral energy sequence.

[0012] Based on the expansion coefficients of each spectral order and the function values ​​of the spectral basis functions at each column index position, a spectral intensity response sequence arranged in the order of column index positions is generated. The average value and standard deviation of the spectral intensity are obtained. The column index positions where the spectral intensity is greater than the sum of the average value and standard deviation are identified as significant regions. These column index positions are recorded to form a significant region index set. The spectral intensity within the significant region index set is averaged to obtain the average spectral intensity of the significant region.

[0013] An adhesion force evaluation function is constructed based on the total curvature change, total spectral energy, spectral distribution skewness, and kurtosis of the one-dimensional elevation curve, and the adhesion force evaluation value is obtained.

[0014] A normalized adhesion index is constructed based on the adhesion strength assessment value, the average spectral intensity of the significant region, and the preset level threshold. The adhesion strength level is then output according to the interval in which the normalized adhesion index is located.

[0015] Optionally, the step of performing three-dimensional measurement of the contact area between the self-adhesive film material and the target adhesion surface to obtain a contact area height map, and generating a zero-offset height matrix based on the height map, specifically includes:

[0016] After the self-adhesive film material is bonded to the target adhesion surface, a three-dimensional measurement device is used to scan the contact area between the two to obtain the surface height data of each pixel in the contact area, and all pixel height data are arranged into a two-dimensional original height matrix according to the row index and column index.

[0017] The average height of all pixels in the original two-dimensional height matrix is ​​averaged to obtain the global average height covering the entire contact area.

[0018] Subtract the global average height from the height data of each pixel in the original 2D height matrix to obtain the zero-offset height matrix, and maintain the same row and column index relationship as the original height matrix in the zero-offset height matrix.

[0019] Optionally, the step of averaging the zero-offset height matrix along the column direction to obtain the average height sequence of each column and generating a one-dimensional elevation curve specifically includes:

[0020] The average height data of each column in the zero offset height matrix is ​​averaged along the row direction to obtain the average height value of each column, and the average height values ​​of each column are arranged into a column-directed projection curve vector in column index order.

[0021] The column projection curve vector is used as a one-dimensional elevation curve along the column direction of the contact area to obtain a one-dimensional elevation curve composed of the average height values ​​of each column.

[0022] Optionally, the step of performing a three-point difference operation on each intermediate sampling point on the one-dimensional elevation curve to obtain a discrete curvature change sequence, and accumulating the absolute values ​​of the discrete curvature changes to obtain the total curvature change, specifically includes:

[0023] In a one-dimensional elevation curve, for each intermediate sampling point other than the beginning and end, a three-point difference operation is used, taking the current sampling point and its adjacent previous and next sampling points as input, to obtain the discrete curvature change sequence at each intermediate sampling point;

[0024] The absolute values ​​at each sampling point in the discrete curvature change sequence are accumulated to obtain the total discrete curvature change at all computable locations, which is then used as the total curvature change of the one-dimensional elevation curve.

[0025] Optionally, the step of performing spectral expansion on the one-dimensional elevation curve based on a family of non-orthogonal spectral basis functions, obtaining expansion coefficients of each spectral order, and accumulating the squares of the expansion coefficients of each spectral order to obtain the total spectral energy specifically includes:

[0026] Construct a custom family of nonorthogonal spectral basis functions. In the custom family of nonorthogonal spectral basis functions, set a unique order index for each order of spectral basis. In each spectral basis, combine the first sine component and the second cosine component, such that the frequency of the second cosine component is three times the frequency of the first sine component, and the amplitude of the second cosine component is inversely proportional to the frequency control parameter corresponding to the frequency of the second cosine component. Form the nonorthogonal spectral basis function values ​​at each column index position.

[0027] The one-dimensional elevation curve is spectrally expanded on a non-orthogonal family of spectral basis functions. By performing a product operation on the elevation curve values ​​and the function values ​​of each order of spectral basis at all column index positions, and performing a weighted sum operation within the column index position range, a normalization coefficient inversely proportional to the total number of column index positions is introduced to obtain the sequence of expansion coefficients for each spectral order.

[0028] The expansion coefficients of each spectral order are squared, and the squared values ​​of the expansion coefficients of all spectral orders are summed to obtain the total spectral energy.

[0029] Optionally, the statistical analysis of the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order to obtain the mean, standard deviation, and spectral distribution skewness and kurtosis based on the third and fourth order central moments specifically includes:

[0030] Statistical analysis is performed on the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order to obtain the mean and standard deviation of the spectral energy sequence;

[0031] When the standard deviation of the spectral energy sequence is greater than zero, the third central moment is calculated on the spectral energy sequence, and the division operation is performed with the cube of the standard deviation as the denominator to obtain the skewness of the spectral energy distribution. At the same time, the fourth central moment is calculated, and the division operation is performed with the fourth power of the standard deviation as the denominator to obtain the kurtosis of the spectral energy distribution.

[0032] When the standard deviation of the spectral energy sequence is zero, the skewness of the spectral energy distribution is set to zero, and the kurtosis of the spectral energy distribution is set to zero.

[0033] Optionally, the step of generating a spectral intensity response sequence arranged in column index order based on the expansion coefficients of each spectral order and the function values ​​of the spectral basis functions at each column index position, obtaining the average value and standard deviation of the spectral intensity, identifying column index positions where the spectral intensity is greater than the sum of the average value and standard deviation as salient regions, recording these column index positions to form a salient region index set, and averaging the spectral intensities within the salient region index set to obtain the average spectral intensity of the salient regions, specifically includes:

[0034] At each column index position of the one-dimensional elevation curve, the absolute value of the product of the spectral basis function value and the expansion coefficient of each spectral order at the corresponding column index position is taken, and the summation operation is performed over the entire spectral order range to form a spectral intensity response sequence in the order of column index positions.

[0035] Perform statistical analysis on the spectral intensity response sequence to obtain the mean and standard deviation of the spectral intensity;

[0036] In the spectral intensity response sequence, select the column index positions where the spectral intensity is greater than the sum of the average and standard deviation of the spectral intensity. Combine the indices of these column index positions into a salient region index set and record the number of elements in the salient region index set.

[0037] Optionally, the step of constructing an adhesion force evaluation function based on the total curvature change, total spectral energy, spectral distribution skewness, and kurtosis of the one-dimensional elevation curve, and obtaining the adhesion force evaluation value, specifically includes:

[0038] The total curvature change of a one-dimensional elevation curve is multiplied by the total spectral energy to obtain the product result that reflects the degree of coupling between the curvature change intensity and the spectral energy.

[0039] The absolute value of the skewness of the spectral energy distribution is taken, and the skewness after taking the absolute value is added to the kurtosis of the spectral energy distribution and the constant to obtain the value of the denominator.

[0040] The product is used as the numerator, and the denominator is used as the denominator to perform a division operation to obtain the value of the adhesion force evaluation function.

[0041] Optionally, the step of constructing a normalized adhesion index based on the adhesion strength assessment value, the average spectral intensity of the significant region, and a preset level threshold, and outputting the adhesion strength level according to the interval where the normalized adhesion index falls, specifically includes:

[0042] When the number of elements in the significant region index set is greater than zero, the average spectral intensity corresponding to each column index position in the significant region index set is calculated to obtain the average spectral intensity of the significant region; when the number of elements in the significant region index set is equal to zero, the average spectral intensity is taken as the average spectral intensity of the significant region.

[0043] The average spectral intensity of the significant region is cubed. The sum of the adhesion force evaluation function value and the cube of the average spectral intensity of the significant region is used as the denominator of the normalization operation, and the adhesion force evaluation function value is used as the numerator of the normalization operation. If the denominator is greater than zero, a division operation is performed to obtain the normalized adhesion index. If the denominator is equal to zero, the normalized adhesion index is set to zero.

[0044] The first-level threshold is set to one-third, and the second-level threshold is set to two-thirds. The first-level threshold and the second-level threshold correspond to the first and second boundary points of the normalized adhesion index interval, respectively.

[0045] When the value of the normalized adhesion index is greater than or equal to zero and less than or equal to the first level threshold, the first adhesion level is output. When the value of the normalized adhesion index is greater than the first level threshold and less than or equal to the second level threshold, the second adhesion level is output. When the value of the normalized adhesion index is greater than the second level threshold, the third adhesion level is output, and the corresponding adhesion level is output as the pattern recognition result.

[0046] The present invention has the following beneficial effects:

[0047] 1. High-precision three-dimensional measurement technology is introduced into the evaluation of self-adhesive film adhesion. By performing dense pixel-level surface scanning on the contact area after bonding, the absolute height data of each pixel is collected to form a two-dimensional original height matrix. Unlike traditional methods that rely solely on macroscopic mechanical testing, this approach obtains a zero-offset matrix by removing the global average height, eliminating the influence of overall bonding height variations and preserving microscopic features. This provides a reliable foundation for subsequent high-precision curvature and spectral analysis. On the one hand, it achieves non-contact, non-destructive acquisition of microscopic morphology, avoiding additional damage to the material surface; on the other hand, the zero-offset processing ensures comparability between different samples, solving the problem of evaluation result deviations caused by differences in measurement conditions.

[0048] 2. For the zero-offset height matrix, a method of projection averaging along the column direction is proposed. By averaging the height values ​​of each column of pixels along the row direction, lateral noise interference is eliminated, and the two-dimensional matrix is ​​simplified into a one-dimensional elevation curve. This curve intuitively reflects the average topographic undulation of the contact area along the column direction, preserving microscopic details while reducing data dimensionality, facilitating subsequent difference and spectral analysis. The coupling of column projection and elevation curve balances data dimensionality reduction and topographic feature preservation. This simplifies subsequent computation, reduces the computational pressure on massive amounts of data during spectral expansion, and avoids the misleading effect of point-to-point noise on curvature extraction, enhancing the stability of the evaluation results. Unlike traditional methods that directly perform difference or local analysis on the two-dimensional matrix, this scheme achieves efficient and reliable one-dimensional processing while ensuring the representation of microscopic features.

[0049] 3. For one-dimensional elevation curves, this scheme introduces a three-point difference method to calculate discrete curvature changes. By differentiating the heights of three adjacent sampling points, the local second-order changes of the elevation curve are quantified. The absolute curvature changes at all intermediate points are summed to obtain the total curvature change of the entire curve. Using three-point difference instead of simple first-order difference can more accurately capture minute undulations and edge effects, avoiding the loss or misjudgment of curvature information caused by first-order gradient fluctuations. The total curvature change can directly reflect surface roughness and undulation intensity, becoming an important characterization of adhesion performance; at the same time, this method is simple to calculate, highly real-time, and easy to deploy quickly in the field.

[0050] 4. A family of non-orthogonal spectral basis functions was constructed, and the one-dimensional elevation curve was expanded on this family of basis functions to obtain multi-order spectral coefficients. The constructed basis functions are combinations of the first sine component and a cosine component with a frequency three times that of the first sine component. The amplitude of the cosine component is inversely proportional to the control parameter corresponding to its frequency, breaking the traditional "orthogonality, equal frequency, equal amplitude" limitations of orthogonal Fourier basis functions, and enabling more flexible matching of the non-sinusoidal and nonlinear characteristics of surface morphology. Through the multi-frequency mixing form of the non-orthogonal basis, higher-order details in the morphology signal can be captured more comprehensively. The introduction of normalization coefficients ensures comparability under different curve lengths or sampling densities. At the same time, the design of the basis function family enhances the sensitivity of the spectral expansion to local anomalies. Compared with traditional Fourier or wavelet expansion, this scheme achieves a balance between frequency domain resolution and time domain local response capability.

[0051] 5. After obtaining the spectral expansion coefficients, this scheme performs statistical analysis on their squared value sequences, calculating the mean and standard deviation, and extracting distribution skewness and kurtosis based on the third and fourth order central moments. Introducing higher-order statistics into spectral analysis, skewness describes the asymmetry of energy distribution, and kurtosis describes the sharpness of the distribution, comprehensively characterizing the spectral energy distribution pattern. This approach goes beyond focusing solely on a single indicator of total energy or dominant frequency energy, and avoids the blind spots of over-reliance on local frequency domain features. Skewness and kurtosis can distinguish between smooth symmetrical fluctuations and sharp, anomalous fluctuations, providing richer statistical information for adhesion assessment; the standard deviation measures the dispersion of spectral energy, enhancing sensitivity to curve irregularities.

[0052] 6. By multiplying and summing the expansion coefficients of each spectral order with the function values ​​of the spectral basis functions at each column index position, a spectral intensity response sequence is constructed to reflect the strength of the frequency domain response at different positions. By statistically analyzing the mean and standard deviation of this sequence, column index positions exceeding the "mean + standard deviation" threshold are identified as salient regions, and the intensity of these salient regions is averaged. Coupled with spatial location information, spatial labeling of local surface anomalies is achieved; salient region localization can identify potential microscopic adhesion hotspots or uneven bonding areas, providing a basis for refined quality control; the average intensity of salient regions further quantifies the comprehensive response intensity of anomaly regions.

[0053] 7. An adhesion force evaluation function is constructed by fusing four core indicators—total curvature change, total spectral energy, spectral skewness, and kurtosis—through a composite function. This function takes into account the intensity of microscopic morphological fluctuations (curvature change), overall frequency energy (total spectral energy), and distribution morphology (skewness and kurtosis), achieving a multi-dimensional characterization of adhesion force. By combining statistical quantities with physically meaningful curvature and spectral indicators, a unified evaluation model is formed. This model can adapt to different material types and surface characteristics, outputting adhesion force values ​​without additional calibration. Furthermore, this function balances the indicator weights, avoiding misjudgments dominated by a single indicator. Compared to traditional single mechanical experiments or methods using only curvature analysis, this approach provides a more comprehensive and interpretable quantitative method for adhesion force.

[0054] 8. A normalized calculation denominator is constructed based on the adhesion force assessment value and the average spectral intensity of the salient region. This denominator, along with the assessment value, forms the numerator. A preset grade threshold maps the normalized index to discrete adhesion force grades. This scheme innovatively uses the cube of the response intensity of the salient region to enhance the amplification effect on abnormally high responses, and achieves a simple and consistent grade division using fixed cutoff points (one-third, two-thirds). The normalized adhesion index eliminates the influence of different absolute dimensions, making different samples directly comparable; the cutoff point setting is simple and clear, facilitating engineering applications and automated judgment. Compared with traditional fuzzy grading or empirical grading methods, this scheme achieves repeatable and automated grade determination, improving assessment efficiency and reliability. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Example, refer to Figure 1 A method for evaluating the adhesion force of self-adhesive films based on pattern recognition, comprising:

[0058] The contact area between the self-adhesive film material and the target adhesion surface is measured in three dimensions to obtain a height map of the contact area, and a zero-offset height matrix is ​​generated based on the height map.

[0059] The zero-offset height matrix is ​​averaged along the column direction to obtain the average height sequence of each column, and a one-dimensional elevation curve is generated.

[0060] On a one-dimensional elevation curve, a three-point difference operation is performed on each intermediate sampling point to obtain a discrete curvature change sequence, and the absolute values ​​of the discrete curvature changes are accumulated to obtain the total curvature change.

[0061] The spectrum of a one-dimensional elevation curve is expanded based on a family of nonorthogonal spectral basis functions. The expansion coefficients of each spectral order are obtained, and the squares of the expansion coefficients of each spectral order are accumulated to obtain the total spectral energy.

[0062] Statistical analysis was performed on the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order to obtain the mean, standard deviation, spectral distribution skewness and kurtosis based on the third and fourth central moments of the spectral energy sequence.

[0063] Based on the expansion coefficients of each spectral order and the function values ​​of the spectral basis functions at each column index position, a spectral intensity response sequence arranged in the order of column index positions is generated. The average value and standard deviation of the spectral intensity are obtained. The column index positions where the spectral intensity is greater than the sum of the average value and standard deviation are identified as significant regions. These column index positions are recorded to form a significant region index set. The spectral intensity within the significant region index set is averaged to obtain the average spectral intensity of the significant region.

[0064] An adhesion force evaluation function is constructed based on the total curvature change, total spectral energy, spectral distribution skewness, and kurtosis of the one-dimensional elevation curve, and the adhesion force evaluation value is obtained.

[0065] A normalized adhesion index is constructed based on the adhesion strength assessment value, the average spectral intensity of the significant region, and the preset level threshold. The adhesion strength level is then output according to the interval in which the normalized adhesion index is located.

[0066] A precise height map of the contact area is obtained through 3D measurement, and global offset is removed. This solves the data deviation caused by traditional methods that rely solely on a single contact area or inconsistent height measurements in tensile tests, allowing subsequent analysis to be based on real and comparable microscopic morphology data. A one-dimensional elevation curve is generated by averaging the 2D height matrix along its column directions, thus preserving the morphological undulation characteristics while reducing dimensionality, avoiding the influence of high-dimensional noise, and reducing computational complexity. The curvature change is extracted through three-point difference operations and accumulated to obtain the total curvature change, effectively quantifying the intensity of surface undulations and overcoming the problem that traditional methods relying on visual inspection or simple roughness indicators cannot accurately reflect micrometer-level undulations. A custom non-orthogonal spectral basis function family is constructed and its curves are expanded, breaking the limitations of the constant amplitude and orthogonality of conventional Fourier basis functions. This allows for the capture of nonlinear and high-order frequency components, solving the problem of insufficient resolution in frequency analysis of complex surface morphology using existing technologies. For the first time, high-order statistics are used for adhesion performance evaluation by statistically analyzing the average, standard deviation, skewness, and kurtosis of the energy sequence composed of squared spectral coefficients, enabling the quantification of distribution morphology information. Furthermore, by leveraging the spatial coupling of spectral basis functions and expansion coefficients, a spectral intensity response sequence is generated. Based on statistical thresholds, significant regions are identified, accurately locating microscopic adhesion hotspots, thus overcoming the limitation of overall assessments that cannot detect local defects. The aforementioned four types of indicators are integrated into an adhesion force evaluation function, achieving multi-dimensional complementarity and improving the robustness and accuracy of the evaluation. A normalized adhesion index is constructed and output in a hierarchical manner, simplifying complex continuous data into hierarchical results, facilitating rapid on-site judgment and automated control in engineering projects.

[0067] The step of performing three-dimensional measurement of the contact area between the self-adhesive film material and the target adhesion surface to obtain a height map of the contact area, and generating a zero-offset height matrix based on the height map, specifically includes:

[0068] After the self-adhesive film material is bonded to the target adhesion surface, a three-dimensional measurement device is used to scan the contact area between the two to obtain the surface height data of each pixel in the contact area, and all pixel height data are arranged into a two-dimensional original height matrix according to the row index and column index.

[0069] The average height of all pixels in the original two-dimensional height matrix is ​​averaged to obtain the global average height covering the entire contact area.

[0070] Subtract the global average height from the height data of each pixel in the original 2D height matrix to obtain the zero-offset height matrix, and maintain the same row and column index relationship as the original height matrix in the zero-offset height matrix.

[0071] Further specific implementation steps include:

[0072] After the self-adhesive film material comes into contact with the target adhesion surface, a 3D measurement device is used to scan the contact area and collect the surface height information at each pixel to construct a 2D original height matrix. Specifically: ;in, The first in the original height matrix Line number The absolute height value of the column pixels; , These are the row index and column index of the height matrix, respectively; , These represent the total number of rows and columns in the heightmap, respectively.

[0073] The global average height of the original height matrix is ​​calculated as follows: ;in, Original height matrix The global average height of all pixels in the array;

[0074] Subtract the global average height from the original height of each pixel to construct a zero-offset height matrix. Specifically: ;in, The zero offset height matrix Line number The height value of the column pixels.

[0075] The step of averaging the zero-offset height matrix along the column directions to obtain the average height sequence of each column and generating a one-dimensional elevation curve specifically includes:

[0076] The average height data of each column in the zero offset height matrix is ​​averaged along the row direction to obtain the average height value of each column, and the average height values ​​of each column are arranged into a column-directed projection curve vector in column index order.

[0077] The column projection curve vector is used as a one-dimensional elevation curve along the column direction of the contact area to obtain a one-dimensional elevation curve composed of the average height values ​​of each column.

[0078] Further specific implementation steps include:

[0079] The zero-offset height value of each column of pixels is averaged in the row direction to obtain the horizontal projection curve vector. Specifically: , ;in, It is a set of one-dimensional vectors consisting of the average height values ​​of each column; For the zero offset height matrix The Middle The average height of the column;

[0080] One-dimensional elevation curve is obtained: ;in, It is a set of one-dimensional elevation curves.

[0081] The process involves performing a three-point difference operation on each intermediate sampling point on a one-dimensional elevation curve to obtain a discrete curvature change sequence, and then summing the absolute values ​​of the discrete curvature changes to obtain the total curvature change. Specifically, this includes:

[0082] In a one-dimensional elevation curve, for each intermediate sampling point other than the beginning and end, a three-point difference operation is used, taking the current sampling point and its adjacent previous and next sampling points as input, to obtain the discrete curvature change sequence at each intermediate sampling point;

[0083] The absolute values ​​at each sampling point in the discrete curvature change sequence are accumulated to obtain the total discrete curvature change at all computable locations, which is then used as the total curvature change of the one-dimensional elevation curve.

[0084] Further specific implementation steps include:

[0085] In the curve The discrete curvature value of each intermediate point is calculated using the three-point difference method, specifically: , ;in, For elevation curves at index positions The discrete curvature change at a point; This is an index for the location of elevation curves;

[0086] Calculate the total change in curvature of the entire curve: ;in, Indexing at all computable locations The sum of the absolute values ​​of the discrete curvature changes.

[0087] The method of performing spectral expansion on a one-dimensional elevation curve based on a family of non-orthogonal spectral basis functions, obtaining expansion coefficients of each spectral order, and accumulating the squares of each expansion coefficient to obtain the total spectral energy specifically includes:

[0088] Construct a custom family of nonorthogonal spectral basis functions. In the custom family of nonorthogonal spectral basis functions, set a unique order index for each order of spectral basis. In each spectral basis, combine the first sine component and the second cosine component, such that the frequency of the second cosine component is three times the frequency of the first sine component, and the amplitude of the second cosine component is inversely proportional to the frequency control parameter corresponding to the frequency of the second cosine component. Form the nonorthogonal spectral basis function values ​​at each column index position.

[0089] The one-dimensional elevation curve is spectrally expanded on a non-orthogonal family of spectral basis functions. By performing a product operation on the elevation curve values ​​and the function values ​​of each order of spectral basis at all column index positions, and performing a weighted sum operation within the column index position range, a normalization coefficient inversely proportional to the total number of column index positions is introduced to obtain the sequence of expansion coefficients for each spectral order.

[0090] The expansion coefficients of each spectral order are squared, and the squared values ​​of the expansion coefficients of all spectral orders are summed to obtain the total spectral energy.

[0091] Further specific implementation steps include:

[0092] Construct a custom family of nonorthogonal spectral basis functions, specifically as follows: , ;in, The first in the family of spectral basis functions The basis functions at position index The function value at that location; This is the order index of the spectral basis functions; This represents the upper limit of the order of the spectral basis.

[0093] Expanding the elevation curves on the aforementioned spectral basis yields the expansion coefficients: ;in, For the elevation curve at the th spectral basis functions The expansion coefficients on;

[0094] The total spectral energy is calculated as follows: ;in, It is the sum of the squares of the coefficients corresponding to all spectral orders in the spectral expansion.

[0095] The statistical analysis of the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order, to obtain the mean, standard deviation, spectral distribution skewness, and kurtosis based on the third and fourth central moments, specifically includes:

[0096] Statistical analysis is performed on the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order to obtain the mean and standard deviation of the spectral energy sequence;

[0097] When the standard deviation of the spectral energy sequence is greater than zero, the third central moment is calculated on the spectral energy sequence, and the division operation is performed with the cube of the standard deviation as the denominator to obtain the skewness of the spectral energy distribution. At the same time, the fourth central moment is calculated, and the division operation is performed with the fourth power of the standard deviation as the denominator to obtain the kurtosis of the spectral energy distribution.

[0098] When the standard deviation of the spectral energy sequence is zero, the skewness of the spectral energy distribution is set to zero, and the kurtosis of the spectral energy distribution is set to zero.

[0099] Further specific implementation steps include:

[0100] Statistical analysis was performed on the squares of the spectral coefficients, and their mean and standard deviation were calculated respectively. , ;in, For spectral energy sequence The average value; The standard deviation of the spectral energy sequence;

[0101] The skewness and kurtosis of the spectral distribution are calculated separately as follows:

[0102] S501, when season: , ;in, It is a skewness measure of the spectral energy distribution, describing the asymmetry of the distribution with respect to the mean; Kujicic is the kurtosis of the spectral energy distribution, describing the degree of concentration of the distribution;

[0103] S502, when season: , .

[0104] The process involves generating a spectral intensity response sequence arranged in column index order based on the expansion coefficients of each spectral order and the function values ​​of the spectral basis functions at each column index position. The average and standard deviation of the spectral intensity are obtained. Column index positions where the spectral intensity is greater than the sum of the average and standard deviation are identified as salient regions. These column index positions are recorded to form a salient region index set. The average spectral intensity of the salient regions is obtained by averaging the spectral intensities within the salient region index set. Specifically, this includes:

[0105] At each column index position of the one-dimensional elevation curve, the absolute value of the product of the spectral basis function value and the expansion coefficient of each spectral order at the corresponding column index position is accumulated over the entire spectral order range, and the spectral intensity response sequence is formed in the order of column index position.

[0106] Perform statistical analysis on the spectral intensity response sequence to obtain the mean and standard deviation of the spectral intensity;

[0107] In the spectral intensity response sequence, select the column index positions where the spectral intensity is greater than the sum of the average and standard deviation of the spectral intensity. Combine the indices of these column index positions into a salient region index set and record the number of elements in the salient region index set.

[0108] Further specific implementation steps include:

[0109] Calculate each lateral position The spectral intensity response at that location is specifically as follows: , ;in, For location index Spectral intensity response value at;

[0110] Calculate the mean and standard deviation of the spectral intensity separately: , ;in, Spectral intensity sequence The average value; The standard deviation of the spectral intensity sequence;

[0111] Extract the strong response regions of the spectrum and construct a set. ;in, To meet the conditions A set of horizontal position indices; For set The number of elements in the middle.

[0112] The adhesion force evaluation function is constructed based on the total curvature change, total spectral energy, spectral distribution skewness, and kurtosis of the one-dimensional elevation curve to obtain the adhesion force evaluation value. Specifically, this includes:

[0113] The total curvature change of a one-dimensional elevation curve is multiplied by the total spectral energy to obtain the product result that reflects the degree of coupling between the curvature change intensity and the spectral energy.

[0114] The absolute value of the skewness of the spectral energy distribution is taken, and the skewness after taking the absolute value is added to the kurtosis of the spectral energy distribution and the constant to obtain the value of the denominator.

[0115] The product is used as the numerator, and the denominator is used as the denominator to perform a division operation to obtain the value of the adhesion force evaluation function.

[0116] Further specific implementation steps include:

[0117] An adhesion force evaluation function is constructed based on the total curvature change, total spectral energy, spectral skewness, and kurtosis of the one-dimensional elevation curve, and the adhesion force evaluation value is obtained, specifically as follows: ;in, This is the value of the adhesion force evaluation function.

[0118] The process involves constructing a normalized adhesion index based on adhesion strength assessment values, average spectral intensity of significant regions, and a preset level threshold. Adhesion strength levels are then output according to the intervals within which the normalized adhesion index falls. Specifically, this includes:

[0119] When the number of elements in the significant region index set is greater than zero, the average spectral intensity corresponding to each column index position in the significant region index set is calculated to obtain the average spectral intensity of the significant region; when the number of elements in the significant region index set is equal to zero, the average spectral intensity is taken as the average spectral intensity of the significant region.

[0120] The average spectral intensity of the significant region is cubed. The sum of the adhesion force evaluation function value and the cube of the average spectral intensity of the significant region is used as the denominator of the normalization operation, and the adhesion force evaluation function value is used as the numerator of the normalization operation. If the denominator is greater than zero, a division operation is performed to obtain the normalized adhesion index. If the denominator is equal to zero, the normalized adhesion index is set to zero.

[0121] The first-level threshold is set to one-third, and the second-level threshold is set to two-thirds. The first-level threshold and the second-level threshold correspond to the first and second boundary points of the normalized adhesion index interval, respectively.

[0122] When the value of the normalized adhesion index is greater than or equal to zero and less than or equal to the first level threshold, the first adhesion level is output. When the value of the normalized adhesion index is greater than the first level threshold and less than or equal to the second level threshold, the second adhesion level is output. When the value of the normalized adhesion index is greater than the second level threshold, the third adhesion level is output, and the corresponding adhesion level is output as the pattern recognition result.

[0123] Further specific implementation steps include:

[0124] when season ;in, For in set Spectral intensity corresponding to all position indices The average value;

[0125] when season ;

[0126] In obtaining Then, the normalized adhesion index is constructed as follows:

[0127] when season ;in, The normalized adhesion index; for cubed;

[0128] when season ;

[0129] Set level threshold: , ;in, , These are the first-level threshold and the second-level threshold of the normalized adhesion index, respectively.

[0130] And according to The numerical values ​​are classified into adhesion levels, specifically: ;in, This outputs the adhesion level.

[0131] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0132] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for evaluating the adhesion force of self-adhesive films based on pattern recognition, characterized in that, include: The contact area between the self-adhesive film material and the target adhesion surface is measured in three dimensions to obtain a height map of the contact area, and a zero-offset height matrix is ​​generated based on the height map. The zero-offset height matrix is ​​averaged along the column direction to obtain the average height sequence of each column, and a one-dimensional elevation curve is generated. On a one-dimensional elevation curve, a three-point difference operation is performed on each intermediate sampling point to obtain a discrete curvature change sequence, and the absolute values ​​of the discrete curvature changes are accumulated to obtain the total curvature change. The spectrum of a one-dimensional elevation curve is expanded based on a family of nonorthogonal spectral basis functions. The expansion coefficients of each spectral order are obtained, and the squares of the expansion coefficients of each spectral order are accumulated to obtain the total spectral energy. Statistical analysis was performed on the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order to obtain the mean, standard deviation, spectral distribution skewness and kurtosis based on the third and fourth central moments of the spectral energy sequence. Based on the expansion coefficients of each spectral order and the function values ​​of the spectral basis functions at each column index position, a spectral intensity response sequence arranged in the order of column index positions is generated. The average value and standard deviation of the spectral intensity are obtained. The column index positions where the spectral intensity is greater than the sum of the average value and standard deviation are identified as significant regions. These column index positions are recorded to form a significant region index set. The spectral intensity within the significant region index set is averaged to obtain the average spectral intensity of the significant region. An adhesion force evaluation function is constructed based on the total curvature change, total spectral energy, spectral distribution skewness, and kurtosis of the one-dimensional elevation curve, and the adhesion force evaluation value is obtained. A normalized adhesion index is constructed based on the adhesion strength assessment value, the average spectral intensity of the significant region, and the preset level threshold. The adhesion strength level is then output according to the interval in which the normalized adhesion index is located.

2. The method for evaluating the adhesion force of self-adhesive films based on pattern recognition according to claim 1, characterized in that, The step of performing three-dimensional measurement of the contact area between the self-adhesive film material and the target adhesion surface to obtain a height map of the contact area, and generating a zero-offset height matrix based on the height map, specifically includes: After the self-adhesive film material is bonded to the target adhesion surface, a three-dimensional measurement device is used to scan the contact area between the two to obtain the surface height data of each pixel in the contact area, and all pixel height data are arranged into a two-dimensional original height matrix according to the row index and column index. The average height of all pixels in the original two-dimensional height matrix is ​​averaged to obtain the global average height covering the entire contact area. Subtract the global average height from the height data of each pixel in the original 2D height matrix to obtain the zero-offset height matrix, and maintain the same row and column index relationship as the original height matrix in the zero-offset height matrix.

3. The method for evaluating the adhesion force of self-adhesive films based on pattern recognition according to claim 2, characterized in that, The step of averaging the zero-offset height matrix along the column directions to obtain the average height sequence of each column and generating a one-dimensional elevation curve specifically includes: The average height data of each column in the zero offset height matrix is ​​averaged along the row direction to obtain the average height value of each column, and the average height values ​​of each column are arranged into a column-directed projection curve vector in column index order. The column projection curve vector is used as a one-dimensional elevation curve along the column direction of the contact area to obtain a one-dimensional elevation curve composed of the average height values ​​of each column.

4. The method for evaluating the adhesion force of self-adhesive films based on pattern recognition according to claim 3, characterized in that, The process involves performing a three-point difference operation on each intermediate sampling point on a one-dimensional elevation curve to obtain a discrete curvature change sequence, and then summing the absolute values ​​of the discrete curvature changes to obtain the total curvature change. Specifically, this includes: In a one-dimensional elevation curve, for each intermediate sampling point other than the beginning and end, a three-point difference operation is used, taking the current sampling point and its adjacent previous and next sampling points as input, to obtain the discrete curvature change sequence at each intermediate sampling point; The absolute values ​​at each sampling point in the discrete curvature change sequence are accumulated to obtain the total discrete curvature change at all computable locations, which is then used as the total curvature change of the one-dimensional elevation curve.

5. The method for evaluating the adhesion force of self-adhesive films based on pattern recognition according to claim 4, characterized in that, The method of performing spectral expansion on a one-dimensional elevation curve based on a family of non-orthogonal spectral basis functions, obtaining expansion coefficients of each spectral order, and accumulating the squares of each expansion coefficient to obtain the total spectral energy specifically includes: Construct a custom family of nonorthogonal spectral basis functions. In the custom family of nonorthogonal spectral basis functions, set a unique order index for each order of spectral basis. In each spectral basis, combine the first sine component and the second cosine component, such that the frequency of the second cosine component is three times the frequency of the first sine component, and the amplitude of the second cosine component is inversely proportional to the frequency control parameter corresponding to the frequency of the second cosine component. Form the nonorthogonal spectral basis function values ​​at each column index position. The one-dimensional elevation curve is spectrally expanded on a non-orthogonal family of spectral basis functions. By performing a product operation on the elevation curve values ​​and the function values ​​of each order of spectral basis at all column index positions, and performing a weighted sum operation within the column index position range, a normalization coefficient inversely proportional to the total number of column index positions is introduced to obtain the sequence of expansion coefficients for each spectral order. The expansion coefficients of each spectral order are squared, and the squared values ​​of the expansion coefficients of all spectral orders are summed to obtain the total spectral energy.

6. The method for evaluating the adhesion force of self-adhesive films based on pattern recognition according to claim 5, characterized in that, The statistical analysis of the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order, to obtain the mean, standard deviation, spectral distribution skewness, and kurtosis based on the third and fourth central moments, specifically includes: Statistical analysis is performed on the spectral energy sequence composed of the squares of the expansion coefficients of each spectral order to obtain the mean and standard deviation of the spectral energy sequence; When the standard deviation of the spectral energy sequence is greater than zero, the third central moment is calculated on the spectral energy sequence, and the division operation is performed with the cube of the standard deviation as the denominator to obtain the skewness of the spectral energy distribution. At the same time, the fourth central moment is calculated, and the division operation is performed with the fourth power of the standard deviation as the denominator to obtain the kurtosis of the spectral energy distribution. When the standard deviation of the spectral energy sequence is zero, the skewness of the spectral energy distribution is set to zero, and the kurtosis of the spectral energy distribution is set to zero.

7. The method for evaluating the adhesion force of a self-adhesive membrane based on pattern recognition according to claim 6, characterized in that, The process involves generating a spectral intensity response sequence arranged in column index order based on the expansion coefficients of each spectral order and the function values ​​of the spectral basis functions at each column index position. The average and standard deviation of the spectral intensity are obtained. Column index positions where the spectral intensity is greater than the sum of the average and standard deviation are identified as salient regions. These column index positions are recorded to form a salient region index set. The average spectral intensity of the salient regions is obtained by averaging the spectral intensities within the salient region index set. Specifically, this includes: At each column index position of the one-dimensional elevation curve, the absolute value of the product of the spectral basis function value and the expansion coefficient of each spectral order at the corresponding column index position is accumulated over the entire spectral order range, and the spectral intensity response sequence is formed in the order of column index position. Perform statistical analysis on the spectral intensity response sequence to obtain the mean and standard deviation of the spectral intensity; In the spectral intensity response sequence, select the column index positions where the spectral intensity is greater than the sum of the average and standard deviation of the spectral intensity. Combine the indices of these column index positions into a salient region index set and record the number of elements in the salient region index set.

8. The method for evaluating the adhesion force of self-adhesive films based on pattern recognition according to claim 7, characterized in that, The adhesion force evaluation function is constructed based on the total curvature change, total spectral energy, spectral distribution skewness, and kurtosis of the one-dimensional elevation curve to obtain the adhesion force evaluation value. Specifically, this includes: The total curvature change of a one-dimensional elevation curve is multiplied by the total spectral energy to obtain the product result that reflects the degree of coupling between the curvature change intensity and the spectral energy. The absolute value of the skewness of the spectral energy distribution is taken, and the skewness after taking the absolute value is added to the kurtosis of the spectral energy distribution and the constant to obtain the value of the denominator. The product is used as the numerator, and the denominator is used as the denominator to perform a division operation to obtain the value of the adhesion force evaluation function.

9. The method for evaluating the adhesion force of self-adhesive films based on pattern recognition according to claim 8, characterized in that, The process involves constructing a normalized adhesion index based on adhesion strength assessment values, average spectral intensity of significant regions, and a preset level threshold. Adhesion strength levels are then output according to the intervals within which the normalized adhesion index falls. Specifically, this includes: When the number of elements in the significant region index set is greater than zero, the average spectral intensity corresponding to each column index position in the significant region index set is calculated to obtain the average spectral intensity of the significant region; when the number of elements in the significant region index set is equal to zero, the average spectral intensity is taken as the average spectral intensity of the significant region. The average spectral intensity of the significant region is cubed. The sum of the adhesion force evaluation function value and the cube of the average spectral intensity of the significant region is used as the denominator of the normalization operation, and the adhesion force evaluation function value is used as the numerator of the normalization operation. If the denominator is greater than zero, a division operation is performed to obtain the normalized adhesion index. If the denominator is equal to zero, the normalized adhesion index is set to zero. The first-level threshold is set to one-third, and the second-level threshold is set to two-thirds. The first-level threshold and the second-level threshold correspond to the first and second boundary points of the normalized adhesion index interval, respectively. When the value of the normalized adhesion index is greater than or equal to zero and less than or equal to the first level threshold, the first adhesion level is output. When the value of the normalized adhesion index is greater than the first level threshold and less than or equal to the second level threshold, the second adhesion level is output. When the value of the normalized adhesion index is greater than the second level threshold, the third adhesion level is output, and the corresponding adhesion level is output as the pattern recognition result.

Images