Tabber glue missing image contrast enhancement method

Abstract

The present application relates to the technical field of image enhancement, in particular to a tab glue missing image contrast enhancement method, comprising the following steps: obtaining a to-be-processed measured tab coating image, combining the high specular reflection characteristics of the tab metal and the diffuse reflection physical characteristics of the coating glue, and calculating the global gray histogram of the measured tab coating image; using a Gaussian mixture model to fit the global gray histogram, determining the background gray peak value, determining the tab metal gray peak value, and determining the tab glue gray peak value. The present application generates an adaptive weight by calculating the local structural similarity between the original image and the preliminary enhanced image, ensures that only the areas with large structural information loss or changes are compensated for details, and the original enhancement effect is maintained in other areas, avoiding problems such as excessive sharpening and noise amplification, and improving the detectability of the missing defect in the machine vision system.

Description

Technical Field

[0001] This invention relates to the field of image enhancement technology, and in particular to a method for enhancing the contrast of images with missing tab adhesive. Background Technology

[0002] Image enhancement technology is a core branch of digital image processing. Its main goal is to improve the visual effect of images by processing them through a series of algorithms.

[0003] The shortcomings of existing technologies lie in the fact that general image enhancement techniques typically employ content-independent processing models, which perform poorly when dealing with industrial images possessing specific physical properties and complex scenes. These techniques aim to universally improve the overall visual effect of images, lacking specific analysis of the optical properties of different materials within the image. For example, when applying global histogram equalization to an image containing large areas of bright metal and localized semi-transparent adhesive coatings, the algorithm's attempt to uniformly distribute the entire grayscale range may compress the grayscale levels of smaller adhesive-coated areas, thereby reducing the subtle differences between missing areas and normally coated areas, making defects even harder to distinguish. Similarly, traditional sharpening operators such as the Laplacian or Sobel operators, while enhancing edges, indiscriminately amplify background noise on metal surfaces, generating numerous artifacts. This is fatal for industrial quality inspection systems requiring high-precision identification, leading to false alarms, reduced production efficiency, and decreased inspection reliability. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a method for enhancing the contrast of images with missing tab adhesive.

[0005] To achieve the above objectives, the present invention employs the following technical solution: a method for enhancing the contrast of an image with missing tab adhesive, comprising the following steps:

[0006] The image of the electrode coating to be processed is obtained, and the global grayscale histogram of the image is calculated by combining the high specular reflection characteristics of the electrode metal and the diffuse reflection physical characteristics of the coating adhesive.

[0007] The global grayscale histogram is fitted using a Gaussian mixture model to determine the background grayscale peak, the tab metal grayscale peak, and the tab adhesive grayscale peak.

[0008] A standardized mapping function is constructed based on the background grayscale peak value, the electrode metal grayscale peak value, the electrode adhesive grayscale peak value, and the theoretical light attenuation physical parameters of the electrode adhesive material.

[0009] The image of the electrode coating to be tested is subjected to contrast enhancement processing using the specified mapping function and the local variance gain term characterizing surface roughness to obtain a first enhanced image;

[0010] Along the physical extension direction of the electrode coating process, the fractional differential amplitude of the electrode coating image under test is calculated to generate a fractional compensation image;

[0011] Calculate the local structural similarity between the electrode coating image to be tested and the first enhanced image, and generate adaptive weights based on the local structural similarity;

[0012] The fractional-order compensation image is fused to the first enhanced image using the adaptive weights to obtain the enhanced result of the electrode coating image under test.

[0013] Preferably, the step of acquiring the image of the tab coating to be processed, and calculating the global grayscale histogram of the image of the tab coating by combining the high specular reflection characteristics of the tab metal and the diffuse reflection physical characteristics of the coating adhesive, specifically includes:

[0014] The original image matrix of the electrode coating to be tested is captured, and the grayscale conversion algorithm is used to perform channel conversion processing on the original image matrix of the electrode coating to be tested. After filtering out background pixels without physical entities using a preset optical edge mask, the electrode coating image to be processed is obtained.

[0015] Traverse all pixels in the image of the electrode coating to be tested, and extract the current gray level value corresponding to each valid pixel in the image of the electrode coating to be tested;

[0016] Construct a grayscale statistics array space containing a preset first number of grayscale levels, and divide the current grayscale value into the corresponding grayscale storage unit within the grayscale statistics array space;

[0017] The total number of pixels contained in each gray level storage unit in the gray level statistical array space is counted to obtain the gray level pixel statistics result, and the total number of effective image pixels contained in the electrode coating image to be tested is obtained.

[0018] The probability of gray level occurrence is calculated by dividing the gray level pixel statistics of each gray level by the total number of effective image pixels.

[0019] A two-dimensional mapping relationship is established based on the gray level occurrence probability corresponding to each gray level. The gray level occurrence probability corresponding to each gray level is plotted into the two-dimensional mapping relationship to generate a global gray level histogram of the electrode coating image to be tested.

[0020] Preferably, the steps of fitting the global grayscale histogram using a Gaussian mixture model to determine the background grayscale peak, the electrode metal grayscale peak, and the electrode adhesive grayscale peak specifically include:

[0021] Obtain a preset Gaussian mixture model, set physical prior parameters based on the pre-calibrated reflectivity of the tab metal material and transmittance of the adhesive material, initialize the model weight parameters of the Gaussian mixture model based on the physical prior parameters, initialize the model mean parameters of the Gaussian mixture model, and initialize the model covariance parameters of the Gaussian mixture model.

[0022] The global grayscale histogram is input into the Gaussian mixture model;

[0023] The first posterior probability is calculated using the expectation-maximization algorithm under the current model weight parameters, model mean parameters, and model covariance parameters.

[0024] The updated model parameters are obtained by updating the model weight parameters, the model mean parameters, and the model covariance parameters based on the first posterior probability.

[0025] The first posterior probability is recalculated using the updated model parameters, and it is determined whether the change in the first posterior probability is less than a preset convergence threshold.

[0026] When the change in the first posterior probability is less than a preset convergence threshold, a converged Gaussian mixture model is obtained.

[0027] Extract the first Gaussian distribution component, the second Gaussian distribution component, and the third Gaussian distribution component contained in the converged Gaussian mixture model.

[0028] Extract the first mean parameter of the first Gaussian distribution component, the second mean parameter of the second Gaussian distribution component, and the third mean parameter of the third Gaussian distribution component;

[0029] A preset grayscale reference range for material reflectance is obtained. The grayscale reference range for material reflectance includes a background range, a metal range, and an adhesive range. The first mean parameter, the second mean parameter, and the third mean parameter are respectively associated and matched with each range in the grayscale reference range for material reflectance. The mean parameter matched to the background range is determined as the background grayscale peak value.

[0030] The mean parameter matched to the metal region is determined as the peak gray value of the tab metal, and the mean parameter matched to the adhesive region is determined as the peak gray value of the tab adhesive.

[0031] Preferably, the step of constructing a standardized mapping function based on the background grayscale peak value, the tab metal grayscale peak value, the tab adhesive grayscale peak value, and the theoretical light attenuation physical parameters of the tab adhesive material specifically includes:

[0032] Obtain all peak data sequences in the global grayscale histogram, and calculate the bandwidth value corresponding to each peak object in the all peak data sequences;

[0033] The bandwidth value is compared with a preset bandwidth threshold, and the peak objects whose bandwidth value is less than the preset bandwidth threshold are marked as high-frequency peaks of metal scratches.

[0034] The high-frequency peaks of the metal scratches in the global grayscale histogram are suppressed using a smoothing filter to obtain suppressed histogram data, and a preset smoothing compensation constant is obtained.

[0035] Calculate the absolute value of the grayscale distance between the grayscale peak value of the tab metal and the grayscale peak value of the tab adhesive;

[0036] Based on the absolute value of the grayscale distance, the mapping grayscale adjustment range between the grayscale peak value of the tab metal and the grayscale peak value of the tab adhesive is calculated using a preset proportional coefficient to obtain the corrected grayscale distribution range.

[0037] An initial mapping curve relationship is constructed based on the corrected grayscale distribution range and the suppressed histogram data.

[0038] A standardized mapping function is generated based on the initial mapping curve relationship and the theoretical light attenuation physical parameters of the tab material.

[0039] Preferably, the step of performing contrast enhancement processing on the electrode coating image under test using the specified mapping function and the local variance gain term characterizing surface roughness to obtain a first enhanced image specifically includes:

[0040] Extract the spatial coordinates of the target pixel in the electrode coating image to be tested;

[0041] A local sliding calculation window is constructed with the aforementioned spatial coordinates as the center.

[0042] Obtain the first grayscale value of all neighboring pixels within the local sliding calculation window;

[0043] Calculate the local grayscale average of the first grayscale value of all the neighboring pixels;

[0044] The local variance value of the local sliding calculation window is calculated based on the local grayscale average value, and the local variance value is used to characterize the local roughness of the electrode coating surface.

[0045] Obtain the preset surface scattering compensation coefficient based on the local roughness and the preset maximum variance value of the local sliding calculation window, and calculate the local variance gain term based on the local variance value;

[0046] The original grayscale value of the target pixel is input into the defined mapping function to obtain the defined mapped grayscale value;

[0047] The specified mapped grayscale value is multiplied with the local variance gain term, and the product result is truncated using a preset anti-overflow threshold function to obtain the enhanced grayscale value corresponding to the target pixel; wherein, the preset anti-overflow threshold function is used to truncate the product result greater than the preset maximum display grayscale level to the preset maximum display grayscale level.

[0048] The product operation and truncation process are performed on all pixels in the electrode coating image to be tested to obtain the enhanced gray value matrix corresponding to the electrode coating image to be tested.

[0049] The enhanced grayscale matrix is ​​output as the first enhanced image.

[0050] Preferably, the step of calculating the fractional-order differential amplitude of the electrode coating image under test along the physical extension direction of the electrode coating process, and generating a fractional-order compensated image, specifically includes:

[0051] Obtain a preset fractional derivative order value, assign heterogeneous gradient response weights to different directions according to the physical extension direction of the tab coating process, and construct a multi-directional fractional derivative mask operator based on the fractional derivative order value and the gradient response weights. The multi-directional fractional derivative mask operator includes a horizontal mask, a vertical mask, a first diagonal mask, and a second diagonal mask.

[0052] The horizontal mask is used to perform convolution extraction on the electrode coating image to be tested, and a horizontal fractional gradient matrix is ​​obtained.

[0053] The vertical mask is used to perform convolution extraction on the electrode coating image to be tested, and a vertical fractional gradient matrix is ​​obtained.

[0054] The first diagonal mask is used to perform convolution extraction on the electrode coating image to be tested to obtain the first diagonal fractional gradient matrix;

[0055] The second diagonal mask is used to perform convolution extraction on the electrode coating image to be tested to obtain the second diagonal fractional gradient matrix;

[0056] The absolute values ​​of the horizontal fractional gradient matrix, the vertical fractional gradient matrix, the first diagonal fractional gradient matrix, and the second diagonal fractional gradient matrix are summed to obtain the fractional differential magnitude matrix.

[0057] The fractional differential magnitude matrix is ​​subjected to a maximum-minimum value normalization mapping process to obtain the mapped fractional differential magnitude matrix.

[0058] A fractional-order compensated image is generated using the mapped fractional-order differential magnitude matrix.

[0059] Preferably, the step of calculating the local structural similarity between the electrode coating image to be tested and the first enhanced image, and generating adaptive weights based on the local structural similarity, specifically includes:

[0060] Based on the local sliding calculation window, the local grayscale average value is extracted as the first local average brightness value of the electrode coating image to be tested, and the second local average brightness value of the first enhanced image within the local sliding calculation window is calculated.

[0061] The first local standard deviation of the electrode coating image to be tested is obtained by taking the square root of the local variance value, and the second local standard deviation of the first enhanced image within the local sliding calculation window is calculated.

[0062] Calculate the first local covariance value between the image of the electrode coating to be tested and the first enhanced image;

[0063] The local structural similarity index between the electrode coating image to be tested and the first enhanced image is calculated based on the first local average brightness value, the second local average brightness value, the first local standard deviation value, the second local standard deviation value, and the first local covariance value.

[0064] Perform an inversion operation on the local structural similarity index to obtain the basic inversion weight values;

[0065] Obtain a preset weight adjustment coefficient, and calculate adaptive weights based on the basic inverse weight values ​​and the preset weight adjustment coefficients.

[0066] Preferably, the step of fusing the fractional-order compensation image to the first enhanced image using the adaptive weights to obtain the enhanced image of the electrode coating under test specifically includes:

[0067] Obtain the first pixel coordinate information of the fractional-order compensated image and the second pixel coordinate information of the first enhanced image;

[0068] Align the first pixel coordinate information and the second pixel coordinate information to determine the processing position of overlapping pixels;

[0069] Extract the first compensation pixel value of the fractional-order compensation image at the overlapping pixel processing position, extract the second enhancement pixel value of the first enhancement image at the overlapping pixel processing position, and extract the third weight value of the adaptive weight at the overlapping pixel processing position.

[0070] The first compensated pixel value is multiplied by the third weight value to obtain the weighted compensation component.

[0071] The weighted compensation component is added to the second enhanced pixel value, and then truncated using a preset anti-overflow threshold function to obtain the fused pixel result value.

[0072] The overlapping pixel processing positions are traversed and processed to obtain a set of fused pixel result values. The set of fused pixel result values ​​is reconstructed into an image format, and the enhancement result of the electrode coating image under test is output.

[0073] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0074] This invention utilizes a Gaussian mixture model to fit the histogram to determine the grayscale peaks of the background, tab metal, and tab adhesive. This peak localization method based on a statistical model, compared to simple threshold segmentation, can more accurately separate the grayscale distributions corresponding to different materials in the image, and can still stably identify them even under uneven lighting or noise interference. The predefined mapping function constructed on this basis is not a universal curve, but rather integrates the separated grayscale peaks with the theoretical light attenuation physical parameters of the tab adhesive material, achieving nonlinear contrast stretching for specific materials. Simultaneously, a local variance gain term characterizing surface roughness is introduced, which locally and adaptively amplifies the texture details of the image while performing global mapping, achieving a dual enhancement of macroscopic contrast and microscopic detail. Furthermore, the fractional derivative amplitude calculated along the physical extension direction of the tab coating process can extract weak edge and subtle texture information that is difficult to capture with conventional integer derivatives, specifically used to compensate for weak signals caused by missing bonding defects. Finally, adaptive weights are generated by calculating the local structural similarity between the original image and the preliminary enhanced image. This ensures that detail compensation is performed only in areas where structural information is lost or changed significantly, while maintaining the original enhancement effect in other areas. This avoids problems such as over-sharpening and noise amplification, and improves the detectability of missing sticker defects in machine vision systems. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0077] Please see Figure 1 This invention provides a technical solution, a method for enhancing the contrast of an image with missing tab adhesive, comprising the following steps:

[0078] The image of the electrode coating to be processed is obtained, and the global grayscale histogram of the image is calculated by combining the high specular reflection characteristics of the electrode metal and the diffuse reflection physical characteristics of the coating adhesive.

[0079] The global grayscale histogram was fitted using a Gaussian mixture model to determine the background grayscale peak, the tab metal grayscale peak, and the tab adhesive grayscale peak.

[0080] A standardized mapping function is constructed based on the background grayscale peak, the grayscale peak of the tab metal, the grayscale peak of the tab adhesive, and the theoretical light attenuation physical parameters of the tab adhesive material.

[0081] The first enhanced image is obtained by using a defined mapping function and a local variance gain term that characterizes surface roughness to perform contrast enhancement processing on the coated image of the electrode under test;

[0082] Along the physical extension direction of the electrode coating process, calculate the fractional differential amplitude of the electrode coating image to be tested, and generate a fractional compensation image.

[0083] Calculate the local structural similarity between the electrode coating image to be tested and the first enhanced image, and generate adaptive weights based on the local structural similarity;

[0084] The fractional-order compensation image is fused into the first enhanced image using adaptive weights to obtain the enhanced result of the electrode coating image under test.

[0085] In this embodiment, the steps of acquiring the image of the electrode coating to be processed, and calculating the global grayscale histogram of the image of the electrode coating to be processed by combining the high specular reflection characteristics of the electrode metal and the diffuse reflection physical characteristics of the coating adhesive, specifically include: acquiring the original image matrix of the electrode coating to be processed by capturing the image; performing channel conversion processing on the original image matrix of the electrode coating to be processed by a grayscale conversion algorithm; and filtering out background pixels without physical entities by using a preset optical edge mask to obtain the image of the electrode coating to be processed; traversing all pixels in the image of the electrode coating to be processed, and extracting the current grayscale value corresponding to each effective pixel in the image of the electrode coating to be processed; and constructing a grayscale level containing a preset first number of grayscale levels. The gray-level statistical array space is used to divide the current gray-level value into the corresponding gray-level storage units within the gray-level statistical array space; the total number of pixels contained in each gray-level storage unit within the gray-level statistical array space is counted to obtain the gray-level pixel statistics result, and the total number of effective image pixels contained in the electrode coating image to be tested is obtained; the gray-level pixel statistics result of each gray level is divided by the total number of effective image pixels to calculate the gray-level occurrence probability corresponding to each gray level; a two-dimensional mapping relationship is established based on the gray-level occurrence probability corresponding to each gray level, and the gray-level occurrence probability corresponding to each gray level is plotted on the two-dimensional mapping relationship to generate the global gray-level histogram of the electrode coating image to be tested.

[0086] Specifically, the original image matrix of the electrode coating to be tested is obtained, which is a color image with a resolution of 2048x1536 pixels. First, a grayscale conversion algorithm is used to perform channel conversion processing on the original image matrix of the electrode coating to be tested, and a weighted average method is used for calculation. The formula is as follows: Where Y is the converted grayscale value, and R, G, and B are the pixel values ​​of the red, green, and blue channels of the original image, respectively. Then, a preset optical edge mask is used to filter out background pixels without physical entities. The optical edge mask is a binary image generated after offline calibration of the positional relationship between a fixed light source and a camera at the production station. The white areas inside the mask correspond to valid electrode detection areas, and the black areas correspond to invalid background. A pixel-by-pixel AND operation is performed between the mask and the grayscale image to obtain the electrode coating image to be processed. Then, all pixels in the electrode coating image to be processed are traversed, i.e., all pixels from coordinates (0,0) to (2048,1536) that were not filtered out by the mask. The current grayscale value corresponding to each valid pixel in the electrode coating image to be processed is extracted, constructing a grayscale statistical array space containing a preset first number of grayscale levels (specifically 256, corresponding to the grayscale range of 0 to 255). This array space is essentially a... A one-dimensional integer array of length 256 is initialized with all values ​​of 0. Then, the current gray level value of each valid pixel is used as an index to increment the corresponding gray level storage unit within the gray level statistics array space by 1. After traversal, the total number of pixels contained in each gray level storage unit within the gray level statistics array space is counted to obtain the gray level pixel statistics result. Simultaneously, all values ​​in the entire gray level statistics array space are summed to obtain the total number of valid image pixels in the electrode coating image to be tested. The gray level pixel statistics result for each gray level is divided by the total number of valid image pixels to calculate the probability of gray level occurrence for each gray level. Finally, a two-dimensional mapping relationship is established based on the gray level occurrence probabilities for each gray level, using gray levels from 0 to 255 as the horizontal axis and the corresponding gray level occurrence probabilities as the vertical axis. The 256 probability values ​​are plotted in the two-dimensional mapping relationship to generate a global gray level histogram of the electrode coating image to be tested.

[0087] In this embodiment, the steps of fitting the global grayscale histogram with a Gaussian mixture model to determine the background grayscale peak, the grayscale peak of the tab metal, and the grayscale peak of the tab adhesive specifically include: obtaining a preset Gaussian mixture model; setting physical prior parameters based on the pre-calibrated reflectivity of the tab metal material and transmittance of the adhesive material; initializing the model weight parameters, the model mean parameters, and the model covariance parameters of the Gaussian mixture model based on the physical prior parameters; inputting the global grayscale histogram into the Gaussian mixture model; calculating the first posterior probability under the current model weight parameters, model mean parameters, and model covariance parameters using the expectation-maximization algorithm; updating the model weight parameters, model mean parameters, and model covariance parameters according to the first posterior probability to obtain the updated model parameters; recalculating the first posterior probability using the updated model parameters; and determining whether the change in the first posterior probability is less than a preset convergence threshold; in the... When the change in posterior probability is less than a preset convergence threshold, a converged Gaussian mixture model is obtained. The first Gaussian distribution component, the second Gaussian distribution component, and the third Gaussian distribution component contained in the converged Gaussian mixture model are extracted. The first mean parameter of the first Gaussian distribution component, the second mean parameter of the second Gaussian distribution component, and the third mean parameter of the third Gaussian distribution component are extracted. A preset material reflectivity grayscale reference interval is obtained, which includes a background interval, a metal interval, and an adhesive interval. The first mean parameter, the second mean parameter, and the third mean parameter are associated and matched with each interval in the material reflectivity grayscale reference interval, respectively. The mean parameter matched to the background interval is determined as the background grayscale peak value. The mean parameter matched to the metal interval is determined as the tab metal grayscale peak value, and the mean parameter matched to the adhesive interval is determined as the tab adhesive grayscale peak value.

[0088] Specifically, a pre-defined Gaussian mixture model is obtained. Physical prior parameters are set based on pre-calibrated reflectivity of the tab metal and transmittance of the adhesive material. For example, experimental measurements under current lighting conditions show that the average grayscale value of the background area is approximately 30, the metal area is approximately 220, and the adhesive area is approximately 110. Based on these physical prior parameters, the mean parameters of the three components of the Gaussian mixture model are initialized to 30, 220, and 110, respectively. Simultaneously, the initial values ​​of the model covariance parameters are set empirically, for example, all set to 20, indicating an initial rough estimate of the dispersion of the grayscale distribution. The model weight parameters of the three components are also initialized to 1 / 3, indicating that the probability of each component appearing is initially considered equal. Then, the global grayscale histogram obtained in the previous step is used as the observation data and input into the Gaussian mixture model. The expectation-maximization algorithm is used for iterative solution. In the E-step, the first posterior probability of each grayscale data point in the histogram belonging to one of the three Gaussian distribution components is calculated under the current model weight parameters, model mean parameters, and model covariance parameters. In the M-step, the model weight parameters, model mean parameters, and model covariance parameters are updated based on the first posterior probability to obtain the updated model parameters. Specifically, the new mean is the average of all data points weighted according to the posterior probability; the new covariance is the average of the squared differences between all data points and the new mean weighted according to the posterior probability; and the new weights are the posterior probabilities of the corresponding components. The sum of probabilities is divided by the total number of data points. Then, the updated model parameters are used to re-enter the E-step to calculate the first posterior probability. It is then determined whether the change in the first posterior probability is less than a preset convergence threshold. This convergence threshold is set based on the change in the log-likelihood function. For example, if the threshold is set to 0.001, and the absolute value of the difference between the log-likelihood function values ​​of two consecutive iterations is less than 0.001, then convergence is determined. When the change in the first posterior probability is less than the preset convergence threshold, a converged Gaussian mixture model is obtained. The first Gaussian distribution component, the second Gaussian distribution component, and the third Gaussian distribution component contained in the converged Gaussian mixture model are extracted. A Gaussian distribution component is used, and the first, second, and third mean parameters of these three components after convergence are extracted respectively. A preset material reflectance grayscale reference range is obtained. The material reflectance grayscale reference range includes a background range (e.g., 0-70), a metal range (e.g., 180-255), and an adhesive range (e.g., 71-179). The first, second, and third mean parameters are associated and matched with each range in the material reflectance grayscale reference range. The mean parameter matched to the background range is determined as the background grayscale peak value, the mean parameter matched to the metal range is determined as the tab metal grayscale peak value, and the mean parameter matched to the adhesive range is determined as the tab adhesive grayscale peak value.

[0089] In this embodiment, the step of constructing a standardized mapping function based on the background grayscale peak, the tab metal grayscale peak, the tab adhesive grayscale peak, and the theoretical light attenuation physical parameters of the tab adhesive material specifically includes: obtaining all peak data sequences in the global grayscale histogram; calculating the bandwidth value corresponding to each peak object in the all peak data sequences; comparing the bandwidth value with a preset bandwidth threshold; marking peak objects with bandwidth values ​​less than the preset bandwidth threshold as high-frequency peaks of metal scratches; and using a smoothing filter to smooth the high-frequency peaks of metal scratches in the global grayscale histogram. Suppression processing is performed to obtain suppressed histogram data, and a preset smoothing compensation constant is acquired. The absolute value of the grayscale distance between the grayscale peak of the tab metal and the grayscale peak of the tab adhesive is calculated. Based on the absolute value of the grayscale distance, the mapping grayscale adjustment amplitude between the grayscale peak of the tab metal and the grayscale peak of the tab adhesive is calculated using a preset scaling factor to obtain the corrected grayscale distribution range. An initial mapping curve relationship is constructed based on the corrected grayscale distribution range and the suppressed histogram data. A standardized mapping function is generated based on the initial mapping curve relationship and the theoretical light attenuation physical parameters of the tab adhesive material, with the following formula: ;in, This represents the result of the calculation of the defined mapping function. This indicates that the input is a grayscale value. Indicates the peak grayscale value of the electrode metal. This indicates the peak grayscale value of the tab adhesive. Indicates the range of grayscale adjustment. Represents the smoothing compensation constant. This represents the preset threshold for preventing division by zero, specifically for small positive numbers. This represents the theoretical light attenuation physical parameters of the tab material.

[0090] Specifically, to obtain the complete peak data sequence from the global grayscale histogram, firstly, the histogram data is subjected to first-order differencing. The apex positions of all peaks are located by finding the point where the difference sign changes from positive to negative, forming a peak data sequence. Next, the bandwidth value corresponding to each peak object in the complete peak data sequence is calculated. Specifically, this is done by searching outwards from the peak apex until the histogram count value drops to 50% of the peak value. The grayscale coordinates on both sides are recorded, and the difference between the two is the bandwidth of that peak. Then, the bandwidth value is compared with a preset bandwidth threshold, which is obtained based on statistics from a large number of normal images. For example, the peak width of normal adhesive and metal areas is usually greater than 10 grayscale levels, while the peaks formed by noise such as scratches are narrower. Therefore, the bandwidth threshold can be set to 5, and peak objects with bandwidth values ​​less than 5 are marked as high-frequency peaks of metal scratches. Finally, a smoothing filter is used to suppress the high-frequency peaks of metal scratches in the global grayscale histogram. Specifically, a 1x7 Gaussian filter with a standard deviation of 2.0 is used to perform a one-dimensional convolution on the histogram data to obtain the suppressed histogram data. A preset smoothing compensation constant is then obtained, which is empirically set to adjust the overall brightness, for example, 2. Next, the absolute value of the grayscale distance between the grayscale peak of the tab metal and the grayscale peak of the tab adhesive is calculated. Based on this absolute value, a preset scaling factor is used to calculate the mapping grayscale adjustment range between the grayscale peaks of the tab metal and the tab adhesive. This scaling factor controls the intensity of contrast stretching, for example, 1.5. If the grayscale distance is 100, the adjustment range is 150, resulting in the corrected grayscale distribution range. Then, an initial mapping curve relationship is constructed based on the corrected grayscale distribution range and the suppressed histogram data. Finally, a standardized mapping function is generated based on the initial mapping curve relationship and the theoretical light attenuation physical parameters of the tab adhesive material. This function is based on the Beer-Lambert law, and its specific form is as follows: ,in, This indicates that the defined mapping function applies to the input grayscale value. The calculation results This represents the grayscale adjustment range obtained from the aforementioned calculation. This represents the theoretical light attenuation physical parameter of the tab material. This is an inherent property of the material and is provided by the supplier or determined experimentally. For example, it is taken as 0.75. This represents the aforementioned determined grayscale peak value of the tab adhesive. This represents the aforementioned peak value of the electrode metal grayscale. This represents a preset threshold for preventing division by zero of tiny positive numbers, typically set to a very small value such as 1e-6. This represents the preset smoothing compensation constant obtained above.

[0091] In this embodiment, the step of performing contrast enhancement processing on the electrode coating image under test using a defined mapping function and a local variance gain term characterizing surface roughness to obtain a first enhanced image specifically includes: extracting the spatial coordinates of the target pixel in the electrode coating image under test; constructing a local sliding calculation window with the spatial coordinates as the center; obtaining the first gray value of all neighboring pixels within the local sliding calculation window; calculating the local gray value average of the first gray value of all neighboring pixels; calculating the local variance value of the local sliding calculation window based on the local gray value average, and using the local variance value to characterize the local roughness of the electrode coating surface; obtaining a preset surface scattering compensation coefficient based on local roughness and a preset maximum variance value of the local sliding calculation window, and calculating the local variance gain term based on the local variance value, with the formula as follows: ;in, This represents the local variance gain term. This represents the local variance value. This indicates the preset maximum variance value. The surface scattering compensation coefficient is represented; the original grayscale value of the target pixel is input into the specified mapping function to obtain the specified mapped grayscale value; the specified mapped grayscale value is multiplied with the local variance gain term, and the product result is truncated using a preset anti-overflow threshold function to obtain the enhanced grayscale value corresponding to the target pixel; wherein, the preset anti-overflow threshold function is used to truncate the product result greater than the preset maximum display grayscale level to the preset maximum display grayscale level; all pixels in the electrode coating image to be tested are traversed to perform product operation and truncation processing to obtain the enhanced grayscale value matrix corresponding to the electrode coating image to be tested; the enhanced grayscale value matrix is ​​output as the first enhanced image.

[0092] Specifically, the spatial coordinates of the target pixel in the image of the electrode coating to be tested are extracted. For example, starting from the first pixel (0,0) at the top left corner of the image, a 5x5 local sliding calculation window is constructed with this spatial coordinate as the center. This window covers the target pixel and its 24 surrounding neighboring pixels. Then, the first grayscale value of all 25 neighboring pixels within the local sliding calculation window is obtained, and the local grayscale average value of the first grayscale value of these 25 pixels is calculated. That is, the sum of all grayscale values ​​is divided by 25. Based on this local grayscale average value, the local variance value of the local sliding calculation window is further calculated, that is, the grayscale value of each pixel is calculated. The average of the sum of squares of the differences between the value and the local grayscale average is used to characterize the local roughness of the electrode coating surface using the local variance value. Then, a preset surface scattering compensation coefficient based on local roughness and a preset maximum variance value of the local sliding calculation window are obtained. The maximum variance value is calculated based on the theoretical limit of 8-bit grayscale images, that is, when half of the pixels in the window are 0 and half are 255, the variance is the largest, approximately 65025. The surface scattering compensation coefficient is an empirical parameter used to adjust the intensity of local contrast enhancement. Based on experimental results, it is set to 1.2, for example. Then, the local variance gain term is calculated based on the local variance value. The calculation method is as follows: ,in, This represents the calculated local variance gain term. This represents the local variance value calculated for the current window. This represents the preset surface scattering compensation coefficient. This indicates the preset maximum variance value. The natural logarithm operation is then performed. The original grayscale value of the target pixel is input into the predefined mapping function constructed in the previous step to obtain the predefined mapped grayscale value. This predefined mapped grayscale value is then multiplied by the previously calculated local variance gain term. A preset anti-overflow threshold function is used to truncate the product result. Specifically, this function checks if the product result is greater than the preset maximum display grayscale level, which is 255 for an 8-bit image. If it is greater than 255, the result is set to 255; if it is less than 0, it is set to 0; otherwise, it remains unchanged. This yields the enhanced grayscale value corresponding to the target pixel (0,0). Finally, the local sliding calculation window is moved with a step size of 1, traversing all pixels in the electrode coating image under test. The multiplication and truncation operations are repeated for each pixel to obtain the enhanced grayscale value matrix corresponding to the electrode coating image under test. This matrix is ​​then output as the first enhanced image.

[0093] In this embodiment, the step of calculating the fractional-order differential magnitude of the electrode coating image to be tested along the physical extension direction of the electrode coating process and generating a fractional-order compensation image specifically includes: obtaining a preset fractional-order differential order value; assigning heterogeneous gradient response weights to different directions according to the physical extension direction of the electrode coating process; constructing a multi-directional fractional-order differential mask operator based on the fractional-order differential order value and the gradient response weights, the multi-directional fractional-order differential mask operator including a horizontal mask, a vertical mask, a first diagonal mask, and a second diagonal mask; performing convolution extraction operations on the electrode coating image to be tested using the horizontal mask to obtain a horizontal fractional-order gradient matrix; and performing convolution operations on the electrode coating image to be tested using the vertical mask. Extraction operations are performed to obtain the vertical fractional gradient matrix; convolution extraction operations are performed on the image of the electrode coating to be tested using the first diagonal mask to obtain the first diagonal fractional gradient matrix; convolution extraction operations are performed on the image of the electrode coating to be tested using the second diagonal mask to obtain the second diagonal fractional gradient matrix; the sum of the absolute values ​​of the horizontal fractional gradient matrix, the vertical fractional gradient matrix, the first diagonal fractional gradient matrix, and the second diagonal fractional gradient matrix is ​​calculated to obtain the fractional differential magnitude matrix; the fractional differential magnitude matrix is ​​subjected to maximum and minimum value normalization mapping to obtain the mapped fractional differential magnitude matrix; and a fractional compensation image is generated using the mapped fractional differential magnitude matrix.

[0094] Specifically, a preset fractional-order differential order value is obtained. The order is a key parameter controlling the characteristics of the differential operator, and its value is between 0 and 1. According to extensive experimental verification, setting the order to 0.5 yields better results for extracting weak edges of the tab coating image. Simultaneously, heterogeneous gradient response weights are assigned to different directions based on the physical extension direction of the tab coating process. For example, if the coating process is performed horizontally, the defect edges in the horizontal direction are more representative, so the gradient response weight in the horizontal direction can be set to 1.5, while the weights in the vertical and diagonal directions can be set to 1.0. Based on the fractional-order differential order value of 0.5 and the aforementioned gradient response weights, a multi-directional fractional-order differential mask operator is constructed using the Grunwald-Letnikov definition. This operator is a set of convolution kernels, including a horizontal mask, a vertical mask, a first diagonal mask, and a second diagonal mask. The size of each mask is typically 5x5 or 7x7. The coefficients are calculated from the fractional order and gamma function, and multiplied by the corresponding directional weights. Then, a horizontal mask is used to perform convolution extraction on the image of the electrode coating to be tested. This involves covering the image with a sliding window and calculating the weighted sum of the pixels within the window and the mask coefficients to obtain the horizontal fractional gradient matrix. Similarly, the vertical, first diagonal, and second diagonal masks are used to perform convolution extraction on the image of the electrode coating to be tested, respectively, to obtain the vertical, first diagonal, and second diagonal fractional gradient matrices. Then, the absolute values ​​of these four gradient matrices are summed. At each pixel location, the absolute values ​​of the corresponding elements of the four matrices are added to obtain a fractional differential magnitude matrix. Each value in this matrix represents the overall edge intensity of that pixel. Finally, the fractional differential magnitude matrix undergoes a maximum-minimum normalization mapping process. First, the maximum value in the entire matrix is ​​found. and minimum value Then for each pixel value Execute mapping The gradient magnitude is linearly mapped to the gray range of 0-255 to obtain the mapped fractional differential magnitude matrix. Finally, the mapped fractional differential magnitude matrix is ​​used to generate a fractional compensated image.

[0095] In this embodiment, the step of calculating the local structural similarity between the electrode coating image to be tested and the first enhanced image, and generating adaptive weights based on the local structural similarity, specifically includes: extracting the local grayscale average value as the first local average brightness value of the electrode coating image to be tested based on a local sliding calculation window, and calculating the second local average brightness value of the first enhanced image within the local sliding calculation window; performing a square root operation on the local variance value to obtain the first local standard deviation value of the electrode coating image to be tested, and calculating the second local standard deviation value of the first enhanced image within the local sliding calculation window; calculating the first local covariance value between the electrode coating image to be tested and the first enhanced image; calculating the local structural similarity index between the electrode coating image to be tested and the first enhanced image based on the first local average brightness value, the second local average brightness value, the first local standard deviation value, the second local standard deviation value, and the first local covariance value; performing an inversion operation on the local structural similarity index to obtain the basic inverted weight value; obtaining a preset weight adjustment coefficient, and calculating the adaptive weight based on the basic inverted weight value and the preset weight adjustment coefficient, using the following formula: ;in, Spatial position coordinates are The adaptive weights corresponding to the pixels. Spatial position coordinates are The local structural similarity index corresponding to the pixels. This represents the weighting adjustment coefficient. Represents the horizontal axis coordinate. Represents the vertical axis coordinate.

[0096] Specifically, based on the local sliding calculation window constructed in the previous steps for processing each pixel, the calculated local grayscale average value is directly extracted as the first local average brightness value of the electrode coating image to be tested, without repeated calculation. Simultaneously, the second local average brightness value is calculated within the local sliding calculation window of the first enhanced image at the same location, i.e., the grayscale values ​​of the 25 pixels in the first enhanced image within that window are averaged. Then, the square root of the previously calculated local variance value is performed to obtain the first local standard deviation value of the electrode coating image to be tested. Similarly, the second local standard deviation value of the first enhanced image within the local sliding calculation window is calculated. Afterwards, the first local covariance value between the electrode coating image to be tested and the first enhanced image is calculated. Based on the first local average brightness value and the second local average brightness value... The first local standard deviation, the second local standard deviation, and the first local covariance are used to calculate the local structural similarity index between the original image and the first enhanced image using structural similarity theory. Then, the local structural similarity index calculated for each pixel location is inverted (i.e., 1 is subtracted from the index value) to obtain the basic inversion weight value. A larger value indicates a greater structural difference between the original and enhanced images. A preset weight adjustment coefficient is then obtained; this coefficient is a global scaling factor set empirically to control the intensity of the final fusion. For example, a more pronounced compensation effect can be achieved by setting the coefficient to 2.0, while a softer effect can be achieved by setting it to 0.8. Here, it is set to 1.5. Finally, adaptive weights are calculated based on the basic inversion weight value and the preset weight adjustment coefficient. The specific calculation is as follows: ,in, Spatial position coordinates are The adaptive weights corresponding to the pixels. This represents the local structural similarity index calculated at the same coordinate point, and its value ranges from -1 to 1. This indicates the preset weight adjustment coefficient, for example, 1.5. and These represent the horizontal and vertical coordinates of a pixel, respectively. This calculation assigns higher fusion weights to regions with lower structural similarity.

[0097] In this embodiment, the step of fusing the fractional-order compensation image to the first enhanced image using adaptive weights to obtain the enhancement result of the electrode coating image under test specifically includes: obtaining the first pixel coordinate information of the fractional-order compensation image and the second pixel coordinate information of the first enhanced image; aligning the first pixel coordinate information and the second pixel coordinate information to determine the overlapping pixel processing position; extracting the first compensation pixel value of the fractional-order compensation image at the overlapping pixel processing position, and extracting the second enhanced pixel value of the first enhanced image at the overlapping pixel processing position, and extracting the third weight value of the adaptive weights at the overlapping pixel processing position; multiplying the first compensation pixel value and the third weight value to obtain the weighted compensation component; summing the weighted compensation component and the second enhanced pixel value, and truncating the result using a preset anti-overflow threshold function to obtain the fused pixel result value, as shown in the formula: ;in, The coordinates of the overlapping processing position are: The values ​​of the fused pixel results, The coordinates of the overlapping processing position are: The second enhanced pixel value, The coordinates of the overlapping processing position are: The first compensated pixel value, The coordinates of the overlapping processing position are: The third weight value, Represents the horizontal axis coordinate. Represents the vertical axis coordinate. This indicates the preset maximum display grayscale level; it iterates through all overlapping pixel processing positions to obtain a set of fused pixel result values, reconstructs the fused pixel result value set into an image format, and outputs the enhancement result of the electrode coating image to be tested.

[0098] Specifically, the first pixel coordinate information of the fractional-order compensated image and the second pixel coordinate information of the first enhanced image are obtained. Since both originate from the same original image and have the same size and coordinate system, pixel-level alignment can be directly performed to determine each coordinate point. For overlapping pixel processing locations, the first compensated pixel value of the fractional-order compensated image at that location is extracted, and the second enhanced pixel value of the first enhanced image at the same location is extracted. Furthermore, the corresponding position is extracted from the adaptive weight matrix calculated in the previous step. The third weight value is calculated, and then the first compensation pixel value is multiplied by the third weight value to obtain the weighted compensation component of the pixel. The magnitude of this component is dynamically determined by the degree of structural change in the image. Then, this weighted compensation component is summed with the second enhancement pixel value, and then truncated by a preset anti-overflow threshold function to obtain the final fused pixel result value. This fusion process can be described as follows: ,in, Indicates coordinates as The final fusion result value of the pixels, This represents the second enhanced pixel value of the first enhanced image at that point. This represents the first compensated pixel value of the fractional-order compensated image at that point. This represents the third weight value of the adaptive weight at that point. This indicates the preset maximum grayscale level; for an 8-bit image, this value is 255. The function ensures that the summation result will not exceed the maximum value. In practice, results less than 0 are truncated and set to 0. Finally, a nested loop iterates through all overlapping pixel locations in the image, for example, from... arrive The weighted fusion and truncation operations are repeated for each pixel to obtain a two-dimensional matrix containing the values ​​of all fused pixel results, i.e., the set of fused pixel result values. The data type of this set is converted to an unsigned 8-bit integer, reconstructed into a standard image format, and the output is the final enhancement result of the electrode coating image under test.

[0099] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for enhancing image contrast in cases where tab adhesive is missing, characterized in that, Includes the following steps: The image of the electrode coating to be processed is obtained, and the global grayscale histogram of the image is calculated by combining the high specular reflection characteristics of the electrode metal and the diffuse reflection physical characteristics of the coating adhesive. The global grayscale histogram is fitted using a Gaussian mixture model to determine the background grayscale peak, the tab metal grayscale peak, and the tab adhesive grayscale peak. A standardized mapping function is constructed based on the background grayscale peak value, the electrode metal grayscale peak value, the electrode adhesive grayscale peak value, and the theoretical light attenuation physical parameters of the electrode adhesive material. The image of the electrode coating to be tested is subjected to contrast enhancement processing using the specified mapping function and the local variance gain term characterizing surface roughness to obtain a first enhanced image; Along the physical extension direction of the electrode coating process, the fractional differential amplitude of the electrode coating image under test is calculated to generate a fractional compensation image; Calculate the local structural similarity between the electrode coating image to be tested and the first enhanced image, and generate adaptive weights based on the local structural similarity; The fractional-order compensation image is fused to the first enhanced image using the adaptive weights to obtain the enhanced result of the electrode coating image under test.

2. The method for enhancing image contrast in case of missing tab adhesive as described in claim 1, characterized in that, The steps of acquiring the image of the electrode coating to be processed, and calculating the global grayscale histogram of the image of the electrode coating by combining the high specular reflection characteristics of the electrode metal and the diffuse reflection physical characteristics of the coating adhesive, specifically include: The original image matrix of the electrode coating to be tested is captured, and the grayscale conversion algorithm is used to perform channel conversion processing on the original image matrix of the electrode coating to be tested. After filtering out background pixels without physical entities using a preset optical edge mask, the electrode coating image to be processed is obtained. Traverse all pixels in the image of the electrode coating to be tested, and extract the current gray level value corresponding to each valid pixel in the image of the electrode coating to be tested; Construct a grayscale statistics array space containing a preset first number of grayscale levels, and divide the current grayscale value into the corresponding grayscale storage unit within the grayscale statistics array space; The total number of pixels contained in each gray level storage unit in the gray level statistics array space is counted to obtain the gray level pixel statistics result, and the total number of effective image pixels contained in the electrode coating image to be tested is obtained. The probability of gray level occurrence is calculated by dividing the gray level pixel statistics of each gray level by the total number of effective image pixels. A two-dimensional mapping relationship is established based on the gray level occurrence probability corresponding to each gray level. The gray level occurrence probability corresponding to each gray level is plotted into the two-dimensional mapping relationship to generate a global gray level histogram of the electrode coating image to be tested.

3. The method for enhancing image contrast in case of missing tab adhesive as described in claim 1, characterized in that, The steps of fitting the global grayscale histogram using a Gaussian mixture model to determine the background grayscale peak, the electrode metal grayscale peak, and the electrode adhesive grayscale peak specifically include: Obtain a preset Gaussian mixture model, set physical prior parameters based on the pre-calibrated reflectivity of the tab metal material and transmittance of the adhesive material, initialize the model weight parameters of the Gaussian mixture model based on the physical prior parameters, initialize the model mean parameters of the Gaussian mixture model, and initialize the model covariance parameters of the Gaussian mixture model. The global grayscale histogram is input into the Gaussian mixture model; The first posterior probability is calculated using the expectation-maximization algorithm under the current model weight parameters, model mean parameters, and model covariance parameters. The updated model parameters are obtained by updating the model weight parameters, the model mean parameters, and the model covariance parameters based on the first posterior probability. The first posterior probability is recalculated using the updated model parameters, and it is determined whether the change in the first posterior probability is less than a preset convergence threshold. When the change in the first posterior probability is less than a preset convergence threshold, a converged Gaussian mixture model is obtained. Extract the first Gaussian distribution component, the second Gaussian distribution component, and the third Gaussian distribution component contained in the converged Gaussian mixture model. Extract the first mean parameter of the first Gaussian distribution component, the second mean parameter of the second Gaussian distribution component, and the third mean parameter of the third Gaussian distribution component; A preset grayscale reference range for material reflectance is obtained. The grayscale reference range for material reflectance includes a background range, a metal range, and an adhesive range. The first mean parameter, the second mean parameter, and the third mean parameter are respectively associated and matched with each range in the grayscale reference range for material reflectance. The mean parameter matched to the background range is determined as the background grayscale peak value. The mean parameter matched to the metal region is determined as the peak gray value of the tab metal, and the mean parameter matched to the adhesive region is determined as the peak gray value of the tab adhesive.

4. The method for enhancing image contrast in case of missing tab adhesive as described in claim 1, characterized in that, The step of constructing a standardized mapping function based on the background grayscale peak value, the electrode metal grayscale peak value, the electrode adhesive grayscale peak value, and the theoretical light attenuation physical parameters of the electrode adhesive material specifically includes: Obtain all peak data sequences in the global grayscale histogram, and calculate the bandwidth value corresponding to each peak object in the all peak data sequences; The bandwidth value is compared with a preset bandwidth threshold, and the peak objects whose bandwidth value is less than the preset bandwidth threshold are marked as high-frequency peaks of metal scratches. The high-frequency peaks of the metal scratches in the global grayscale histogram are suppressed using a smoothing filter to obtain suppressed histogram data, and a preset smoothing compensation constant is obtained. Calculate the absolute value of the grayscale distance between the grayscale peak value of the tab metal and the grayscale peak value of the tab adhesive; Based on the absolute value of the grayscale distance, the mapping grayscale adjustment range between the grayscale peak value of the tab metal and the grayscale peak value of the tab adhesive is calculated using a preset proportional coefficient to obtain the corrected grayscale distribution range. An initial mapping curve relationship is constructed based on the corrected grayscale distribution range and the suppressed histogram data. A standardized mapping function is generated based on the initial mapping curve relationship and the theoretical light attenuation physical parameters of the tab material.

5. The method for enhancing image contrast in case of missing tab adhesive as described in claim 1, characterized in that, The step of performing contrast enhancement processing on the electrode coating image under test using the specified mapping function and the local variance gain term characterizing surface roughness to obtain the first enhanced image specifically includes: Extract the spatial coordinates of the target pixel in the electrode coating image to be tested; A local sliding calculation window is constructed with the aforementioned spatial coordinates as the center. Obtain the first grayscale value of all neighboring pixels within the local sliding calculation window; Calculate the local grayscale average of the first grayscale value of all the neighboring pixels; The local variance value of the local sliding calculation window is calculated based on the local grayscale average value, and the local variance value is used to characterize the local roughness of the electrode coating surface. Obtain the preset surface scattering compensation coefficient based on the local roughness and the preset maximum variance value of the local sliding calculation window, and calculate the local variance gain term based on the local variance value; The original grayscale value of the target pixel is input into the defined mapping function to obtain the defined mapped grayscale value; The specified mapped grayscale value is multiplied with the local variance gain term, and the product result is truncated using a preset anti-overflow threshold function to obtain the enhanced grayscale value corresponding to the target pixel; wherein, the preset anti-overflow threshold function is used to truncate the product result greater than the preset maximum display grayscale level to the preset maximum display grayscale level. The product operation and truncation process are performed on all pixels in the electrode coating image to be tested to obtain the enhanced gray value matrix corresponding to the electrode coating image to be tested. The enhanced grayscale matrix is ​​output as the first enhanced image.

6. The method for enhancing image contrast in case of missing tab adhesive as described in claim 1, characterized in that, The step of calculating the fractional-order differential amplitude of the electrode coating image under test along the physical extension direction of the electrode coating process, and generating a fractional-order compensated image, specifically includes: Obtain a preset fractional derivative order value, assign heterogeneous gradient response weights to different directions according to the physical extension direction of the tab coating process, and construct a multi-directional fractional derivative mask operator based on the fractional derivative order value and the gradient response weights. The multi-directional fractional derivative mask operator includes a horizontal mask, a vertical mask, a first diagonal mask, and a second diagonal mask. The horizontal mask is used to perform convolution extraction on the electrode coating image to be tested, and a horizontal fractional gradient matrix is ​​obtained. The vertical mask is used to perform convolution extraction on the electrode coating image to be tested, and a vertical fractional gradient matrix is ​​obtained. The first diagonal mask is used to perform convolution extraction on the electrode coating image to be tested to obtain the first diagonal fractional gradient matrix; The second diagonal mask is used to perform convolution extraction on the electrode coating image to be tested to obtain the second diagonal fractional gradient matrix; The absolute values ​​of the horizontal fractional gradient matrix, the vertical fractional gradient matrix, the first diagonal fractional gradient matrix, and the second diagonal fractional gradient matrix are summed to obtain the fractional differential magnitude matrix. The fractional differential magnitude matrix is ​​subjected to a maximum-minimum value normalization mapping process to obtain the mapped fractional differential magnitude matrix. A fractional-order compensated image is generated using the mapped fractional-order differential magnitude matrix.

7. The method for enhancing image contrast in case of missing tab adhesive as described in claim 5, characterized in that, The step of calculating the local structural similarity between the electrode coating image to be tested and the first enhanced image, and generating adaptive weights based on the local structural similarity, specifically includes: Based on the local sliding calculation window, the local grayscale average value is extracted as the first local average brightness value of the electrode coating image to be tested, and the second local average brightness value of the first enhanced image within the local sliding calculation window is calculated. The first local standard deviation of the electrode coating image to be tested is obtained by taking the square root of the local variance value, and the second local standard deviation of the first enhanced image within the local sliding calculation window is calculated. Calculate the first local covariance value between the image of the electrode coating to be tested and the first enhanced image; The local structural similarity index between the electrode coating image to be tested and the first enhanced image is calculated based on the first local average brightness value, the second local average brightness value, the first local standard deviation value, the second local standard deviation value, and the first local covariance value. Perform an inversion operation on the local structural similarity index to obtain the basic inversion weight values; Obtain a preset weight adjustment coefficient, and calculate adaptive weights based on the basic inverse weight values ​​and the preset weight adjustment coefficients.

8. The method for enhancing image contrast in case of missing tab adhesive as described in claim 1, characterized in that, The step of fusing the fractional-order compensation image to the first enhanced image using the adaptive weights to obtain the enhanced result of the electrode coating image under test specifically includes: Obtain the first pixel coordinate information of the fractional-order compensated image and the second pixel coordinate information of the first enhanced image; Align the first pixel coordinate information and the second pixel coordinate information to determine the processing position of overlapping pixels; Extract the first compensation pixel value of the fractional-order compensation image at the overlapping pixel processing position, extract the second enhancement pixel value of the first enhancement image at the overlapping pixel processing position, and extract the third weight value of the adaptive weight at the overlapping pixel processing position. The first compensated pixel value is multiplied by the third weight value to obtain the weighted compensation component. The weighted compensation component is added to the second enhanced pixel value, and then truncated using a preset anti-overflow threshold function to obtain the fused pixel result value. The overlapping pixel processing positions are traversed and processed to obtain a set of fused pixel result values. The set of fused pixel result values ​​is reconstructed into an image format, and the enhancement result of the electrode coating image under test is output.

Images