Automatic identification and evaluation method for printing chromatic aberration of label tape

Abstract

The invention provides a trademark tape printing color difference automatic identification and evaluation method, which comprises the steps of extracting boundary positions of color layers from a captured trademark multi-color overprinting image, comparing the boundary positions of the color layers with standard overprinting positions, identifying position differences of the color layers, and obtaining overprinting offset of each region; comparing the overprinting offset of each area with a preset threshold value to determine an overprinting precision grade, acquiring a printing process record, extracting the front-layer ink drying time from the printing process record, and performing matching evaluation on the overprinting precision grade and the front-layer ink drying time to obtain a wetting influence degree; and separating front and back color layer spectrum signals from corresponding positions of the marked color mixing area coordinates, and classifying and grouping the front and back color layer spectrum signals according to hues to obtain a chromaticity distribution diagram of each color layer.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to an automatic identification and evaluation method for color difference in trademark printing. Background Technology

[0002] In the modern printing industry, multi-color overprinting technology for trademarks plays a crucial role, directly impacting the visual appeal of product packaging and the transmission of brand image. Especially in demanding commercial environments, accurate color reproduction is not only a technical indicator but also a key factor in market competition. Ensuring color difference control in multi-color overprinting has become a core task of printing quality management; even slight deviations can lead to consumer misjudgment of product quality. However, current multi-color overprinting color difference detection suffers from a significant technical limitation: existing methods often focus on improving the precision of a single step, failing to fully consider the coupling effects between different steps. Particularly when printing involves multiple layers of ink, simply optimizing one step often leads to imbalances in other steps, resulting in poor overall quality. This one-sidedness makes the match between test results and actual printing quality poor, hindering the handling of complex production scenarios. An even more critical issue lies in the wetting and mutual solubility between ink layers, a core factor affecting color difference control. On high-speed rotary production lines for trademark printing, to ensure a production efficiency of hundreds of meters per minute, printing units of each color are arranged continuously. The next layer of ink must be overprinted immediately while the previous layer is still wet. At this point, when the subsequent ink comes into contact with the previous wet film, solvent penetration and color mixing occur, causing ink molecules to diffuse abnormally at the interface, resulting in unintended color transitions. This wetting and mixing is particularly noticeable in practice. For example, when printing a trademark design containing cyan, yellow, magenta, and black, if the yellow ink layer is applied first, followed immediately by the magenta layer, the solvent penetrates from the magenta layer into the yellow wet film. The yellow pigment particles are partially dissolved and mixed with the magenta, resulting in a blurred, purplish-yellow band at what should be a sharp, distinct edge, severely deviating from the pure color overlay effect of the design. The paradox of this phenomenon is particularly thorny. When the registration accuracy is high, the alignment of each color layer is more accurate, and theoretically, the color superposition should be more standardized and closer to the design expectation. However, in reality, due to the significant increase in the contact area between ink layers caused by precise registration, the molecular diffusion channels at the wetting interface increase, making it easier for the later ink layer to penetrate the previous wet film, resulting in an exacerbation of color mixing. Taking red and yellow ink layers as an example, under low-precision registration, although there is a slight misalignment, the contact edges are limited, and color mixing is limited to small blurry spots. Once the registration accuracy is improved to high precision, such as when the error is controlled within 0.1 mm, the tight adhesion of the complete layers will amplify the wetting and mixing effect, forming a clear orange transition band at the interface, which may even extend inward by several millimeters, causing the color difference of the entire color block to exceed the standard, with the DeltaE value exceeding the acceptable standard of 2.0. This paradox of the higher the precision, the more serious the problem, directly challenges the effectiveness of traditional detection methods.The reason why it's impossible to print the next layer after the previous layer of ink is completely dry is that label tape printing typically uses rotary high-speed production lines designed for continuous operation. Adding drying time between layers would cause production line downtime, reducing production efficiency from tens of thousands of meters per hour to less than half. It would also increase the size and energy consumption of drying equipment, making it unsuitable for large-volume commercial printing demands, such as urgent orders for holiday promotional packaging. Furthermore, the drying time of water-based or solvent-based inks commonly used in flexographic printing often takes several seconds to tens of seconds, which is incompatible with the production line speed, further amplifying the risk of wetting and mixing. In high-humidity workshop environments, this phenomenon is exacerbated by the decrease in ink viscosity, expanding the color mixing range to affect the entire printing area. How to effectively identify and control unintended color mixing caused by wetting and mixing, especially the aggravating effect under high-precision registration, has become a key business challenge in multi-color printing color difference detection, directly hindering the upgrade of label tape printing to high-quality standards. Summary of the Invention

[0003] This invention provides an automatic identification and evaluation method for color difference in printing on trademarks, mainly including: Extract the boundary positions of each color layer from the captured multi-color overprinted image of the trademark, compare the boundary positions of each color layer with the standard overprinting position, and obtain the overprinting offset of each region. The overprinting accuracy level is determined based on the overprinting offset of each region. The drying time of the previous layer ink is extracted from the printing process record. The overprinting accuracy level is matched with the drying time of the previous layer ink to obtain the degree of wetting effect. Spectral features are extracted from the chromaticity values ​​of the interface region between the front and rear color layers to obtain hue shift values. The hue shift values ​​are compared with a dynamic threshold based on the degree of wetting influence to mark the color mixing region and output the region coordinates. The contact area between the front and rear color layers is extracted from the region coordinates. The spectral signals of the front and rear color layers are separated from the corresponding positions in the color mixing region, and grouped according to hue to obtain the color distribution map of each color layer; A comprehensive evaluation is performed based on the color distribution diagrams of each color layer to obtain the degree of color mixing at the interface. The degree of color mixing at the interface is compared with the adjustment threshold based on the overprinting accuracy level, and an indicator of increased color difference is output. The color difference increase indicator and the overprint offset are matched and categorized according to their regional positions to generate an overprint quality assessment report.

[0004] Furthermore, the step of extracting the boundary positions of each color layer from the captured multi-color overprinted image of the trademark, comparing the boundary positions of each color layer with the standard overprinting position, and obtaining the overprinting offset of each region includes: performing edge gradient detection on the multi-color overprinted image of the trademark using the Sobel operator, obtaining the edge gradient intensity value of each color layer, constructing a boundary contour set based on pixels whose gradient intensity values ​​exceed a preset threshold, filtering closed contours from the boundary contour set and calculating the centroid coordinates of the contours; calculating the Euclidean distance between the centroid coordinates of the contours and the standard overprinting reference coordinates to determine the offset anomaly points, performing spatial clustering of the offset anomaly points using a density clustering method, dividing the overprinting region based on the clustering results and generating a region code; extracting the corner points at the color layer intersections from the image blocks corresponding to the region codes as alignment feature points, fitting the offset vector between the alignment feature points and the standard position using the least squares method, calculating the components of the offset vector in the horizontal and vertical directions, and obtaining the overprinting offset of each region.

[0005] Furthermore, the step of determining the registration accuracy level based on the registration offset of each region, extracting the drying time of the previous layer ink from the printing process record, and matching the registration accuracy level with the drying time of the previous layer ink to obtain the degree of wetting influence includes: comparing the value of the registration offset of each region with the grading threshold to determine a high accuracy level, medium accuracy level, or low accuracy level; reading the completion time stamp of the previous layer ink printing and the start time stamp of the subsequent layer ink printing from the printing equipment process record database, calculating the difference between the completion time stamp of the previous layer ink printing and the start time stamp of the subsequent layer ink printing to obtain the drying time of the previous layer ink; and using the registration accuracy level and the drying time of the previous layer ink as an index to look up the wetting influence mapping table to obtain the wetting influence coefficient as the degree of wetting influence.

[0006] Furthermore, the step of extracting spectral features from the chromaticity values ​​of the interface region between the front and rear color layers to obtain a hue shift value, comparing the hue shift value with a dynamic threshold based on the degree of wetting influence, marking the color mixing region and outputting the region coordinates, and extracting the contact area between the front and rear color layers from the region coordinates includes: collecting reflectance spectral data of the interface between the front and rear color layers, performing principal component analysis on the reflectance spectral data to extract the first three principal components as spectral feature vectors, calculating the difference between the actual hue angle and the standard hue angle to obtain the hue shift value; calculating a dynamic threshold based on the degree of wetting influence, marking pixels with hue shift values ​​exceeding the dynamic threshold with eight-neighbor connected components to form a color mixing region and outputting a set of coordinates for the color mixing region; extracting the vertex coordinates of the minimum bounding rectangle from the boundary pixel coordinates of the color mixing region, calculating the product of the rectangle's length and width to obtain the contact area between the front and rear color layers.

[0007] Furthermore, the step of separating the spectral signals of the front and rear color layers from the corresponding positions of the color mixing region, classifying and grouping them according to hue, and obtaining the chromaticity distribution map of each color layer includes: using an independent component analysis algorithm to perform blind source separation on the original spectral data of the color mixing region; matching the spectral curve after blind source separation with the characteristic spectra of the four-color inks to obtain the spectral signals of the front and rear color layers; converting the spectral signals of the front and rear color layers to the CIE Lab chromaticity space, calculating the hue value, and grouping them according to hue similarity using K-means clustering; statistically analyzing the chromaticity intensity values ​​based on the spatial distribution density of each hue group member, and constructing the chromaticity distribution map of each color layer through two-dimensional interpolation.

[0008] Furthermore, the step of comprehensively evaluating the interface color mixing degree based on the color distribution maps of each color layer, and comparing the interface color mixing degree with the adjustment threshold based on the overprinting accuracy level to output a color difference increase indicator, includes: normalizing the hue shift value, the drying time of the previous layer ink, and the contact area, and then performing a weighted summation to obtain a comprehensive evaluation value; extracting the boundary gradient peak and the center color concentration from the color distribution maps of each color layer, calculating the ratio of the boundary gradient peak and the center color concentration, and statistically analyzing the pixel proportion to obtain the interface color mixing degree; querying the adjustment threshold based on the overprinting accuracy level, and if the product of the interface color mixing degree and the comprehensive evaluation value exceeds the adjustment threshold, then assigning a color difference increase indicator to the abnormal area and recording the location coordinates and color mixing degree value.

[0009] Furthermore, the step of matching and classifying the color difference increase markers and the overprint offsets according to their regional locations to generate an overprint quality assessment report includes: performing spatial matching based on the position coordinates of the color difference increase markers and the regional codes of the overprint offsets, establishing a hash table index, realizing the classification and aggregation of markers and offsets within the same region, and generating a data structure containing region number, number of markers, average offset, and color mixing type as the overprint quality assessment report.

[0010] Furthermore, after generating the overprint quality assessment report, the recognition threshold is optimized and the overprint quality recognition benchmark is updated based on the assessment report. The updated benchmark is then used to perform batch processing on subsequent images, and the final overprint quality assessment result is output.

[0011] Furthermore, the step of optimizing the identification threshold and updating the overprint quality identification benchmark based on the evaluation report, and using the updated benchmark to perform batch processing on subsequent images to output the final overprint quality evaluation result includes: statistically analyzing the density of abnormal markers in the overprint quality evaluation report, adjusting the identification threshold based on the offset characteristics of high-incidence areas to obtain the updated benchmark; using the updated benchmark to extract the overprint offset and interface color mixing degree from subsequent batch images, establishing the correspondence between the overprint offset and the interface color mixing degree, and outputting the final overprint quality evaluation result if the overall color difference variation deviation is within the tolerance range.

[0012] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses an automatic identification and evaluation method for color difference in trademark printing, proposing an integrated solution to address color difference problems caused by registration misalignment, ink wetting effects, and abnormal color mixing in multi-color printing. This invention identifies registration misalignment by extracting the boundary positions of each color layer and comparing them with standard positions, and evaluates the accuracy level based on preset thresholds. Simultaneously, it analyzes the matching evaluation of the wetting effect of the previous ink drying time to quantify the degree of solvent wetting. Further, it extracts hue offset values ​​through spectral features, marks mixed color areas using dynamic threshold comparison, and comprehensively evaluates the degree of interface color mixing by combining contact area and color distribution maps. Finally, it identifies abnormal color mixing and outputs color difference indicators. This invention matches the color difference indicators with the misalignment to generate a quality evaluation report, optimizes the identification threshold, updates the benchmark, batch processes images, and outputs the final evaluation results. This invention achieves end-to-end monitoring and optimization from registration accuracy to color mixing effects, effectively improving the accuracy and consistency of printing quality evaluation. Attached Figure Description

[0013] Figure 1 This is a flowchart of an automatic identification and evaluation method for color difference in trademark printing according to the present invention.

[0014] Figure 2 This is a schematic diagram of an automatic identification and evaluation method for color difference in trademark printing according to the present invention.

[0015] Figure 3 This is another schematic diagram of an automatic identification and evaluation method for color difference in trademark printing according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0017] like Figures 1-3 This embodiment of the automatic identification and evaluation method for color difference in trademark printing may specifically include: S101. Extract the boundary positions of each color layer from the captured multi-color overprinted image of the trademark, compare the boundary positions of each color layer with the standard overprinting position, identify the differences in the position of each color layer, and obtain the overprinting offset of each region.

[0018] The Sobel operator is used to perform edge gradient detection on the multi-color overprinted image of the trademark, obtaining the gradient intensity values ​​of each color layer. A set of boundary contours is constructed based on pixels whose gradient intensity values ​​exceed a preset threshold. Closed contours are selected from this set, and their centroid coordinates are calculated. The Euclidean distance between the centroid coordinates of the contours and pre-established standard overprinting reference coordinates is calculated. If the distance exceeds a preset offset threshold, it is identified as an anomaly. The standard overprinting reference coordinates are derived from the standard overprinted image and established by calculating the centroid coordinates of its closed contours. The DBSCAN density clustering method is used to spatially cluster the anomalies. The input is the anomaly coordinates, and the process includes defining the neighborhood radius ε and the minimum number of points MinPts, identifying core points and expanding the clusters, and outputting cluster labels. Based on the clustering results, the overprinted regions are divided, and region codes are generated. Corner points at the color layer boundaries are extracted from the image blocks corresponding to the region codes as alignment feature points. The Harris corner detection algorithm is used, which involves calculating the image gradient and obtaining the corner response value R = det(M) - k(trace(M)). 2 Where M is the gradient covariance matrix, and k is an empirical constant from 0.04 to 0.06, points whose response values ​​exceed a threshold are selected as corner points. The least squares method is used to fit the offset vector between the aligned feature points and the standard position. The process involves minimizing the sum of squared residuals to solve for the linear model parameters. The input is the feature point coordinates (x...). i y i ) and standard position (x) j y j The offset vector (dx, dy) is fitted, and the overprinting offset of each region is determined by calculating the horizontal component dx and the vertical component dy of the offset vector.

[0019] Specifically, in one implementation, when using the Sobel operator to perform edge gradient detection on a multi-color overprinted trademark image, gradient components are calculated in both the horizontal and vertical directions.

[0020] Specifically, for trademark images with four color layers—cyan, magenta, yellow, and black—the Sobel operator calculates the gradient intensity value at each pixel location using a 3×3 convolution kernel. When the gradient intensity value exceeds a preset threshold, the pixel is marked as an edge point.

[0021] It should be noted that due to the varying ink thicknesses of different color layers in trademark printing, the edge gradient intensity of the cyan layer is typically higher than that of the yellow layer. Therefore, the preset threshold needs to be dynamically adjusted based on different color layers. Optionally, the specific method is as follows: calculate the average ink thickness h of each color layer, and set the threshold T = 0.5 * h, where h is the average thickness. Connect the edge points into closed contours using connected component analysis. Specifically, use 8-connectivity judgment, starting from the edge point and scanning neighboring pixels. If pixel values ​​are similar, assign the same label, traversing row by row until all points are connected to form a closed region. Calculate the arithmetic mean of all pixel coordinates within each closed contour as the contour centroid coordinates, which represent the geometric center position of the color layer.

[0022] For example, the standard overlay reference coordinates are an ideal set of position coordinates established during the initial calibration of the printing equipment. After calculating the centroid of the outline, that is, the Euclidean distance between the center point coordinates extracted from the printing outline by the image processing algorithm and the standard reference, if the offset distance of a centroid exceeds a preset offset threshold, the centroid is marked as an offset anomaly point.

[0023] Preferably, the DBSCAN density clustering method is used to spatially cluster the offset outliers. This method groups spatially adjacent outliers into the same cluster by setting two parameters: neighborhood radius and minimum number of points.

[0024] For example, when a trademark tape is printed at high speed, the misregistration caused by uneven tension often exhibits regional characteristics. The direction and magnitude of the misregistration in the left region are significantly different from those in the right region. Density clustering can classify these outliers with similar misregistration characteristics into the same region and assign a unique code identifier to each region.

[0025] In one possible implementation, when extracting corner points at the color layer boundaries from the image patch corresponding to the region encoding, the Harris corner detection algorithm is used to identify feature points with significant gradient changes. These corner points are typically located at the edges of different color layers, such as the overlapping boundary between magenta and yellow. The corresponding feature points are linearly fitted to standard positions using the least squares method to construct an overdetermined system of equations to solve for the optimal estimate of the offset vector.

[0026] Specifically, the offset vector is decomposed into a horizontal component and a vertical component. The horizontal component reflects the registration deviation of the trademark strip in the transverse direction of the printing cylinder, and the vertical component reflects the alignment error in the longitudinal paper feeding process. The combination of the two components is the registration offset of the area.

[0027] S102. Determine the registration accuracy level by comparing the registration offset of each region with the preset threshold, obtain the printing process record and extract the drying time of the previous layer ink from it, and match and evaluate the registration accuracy level with the drying time of the previous layer ink to obtain the degree of wetting effect.

[0028] The numerical range of the overprinting offset in each region is compared with preset grading thresholds. If the offset is less than the first preset threshold, it is determined to be a high-precision level; if the offset is between the first and second preset thresholds, it is determined to be a medium-precision level; and if the offset exceeds the second preset threshold, it is determined to be a low-precision level. The timestamps of the completion of the previous layer ink printing and the start of the subsequent layer ink printing are read from the printing equipment's process record database, and the difference between the two timestamps is calculated to obtain the drying time of the previous layer ink. Using a pre-established wetting effect mapping table, with the overprinting accuracy level as the first dimension index and the previous layer ink drying time as the second dimension index, the corresponding wetting effect coefficient is obtained. The wetting effect coefficient reflects the degree of solubilization and penetration of the subsequent ink into the previous wet film, thus determining the degree of wetting effect.

[0029] Specifically, in one implementation, the classification of overprint accuracy levels is based on the quality control standards of the trademark tape printing industry.

[0030] Specifically, the first and second preset thresholds are dynamically adjusted based on the ink type and printing material. For water-based inks printed on coated paper, the thresholds are relatively small, while for solvent-based inks printed on film, the thresholds are correspondingly wider. The printing process record database stores the real-time operating parameters of each printing unit. The timestamp of the completion of the previous ink layer printing is triggered and recorded by a photoelectric sensor on the printing cylinder. When the label tape passes through the printing unit, the sensor detects that the printing cylinder has rotated to the marked position and generates a timestamp. The timestamp of the start of the subsequent ink layer printing is obtained using the same mechanism. The difference between the two timestamps represents the exposure time of the previous ink layer in the air, which directly affects the degree of solvent evaporation from the ink surface.

[0031] Preferably, the wetting effect mapping table is a two-dimensional lookup table established using a large amount of experimental data.

[0032] In one possible implementation, the construction of the mapping table involves conducting printing experiments under different overprinting accuracy levels and different drying times. The absorption spectrum between ink layers is measured using a spectrometer in a specific wavelength range, such as 400 to 700 nanometers, to indirectly estimate the solvent migration amount S, where S is the solvent migration amount in percentage. The measured values ​​are then normalized and stored in the table as a wetting effect coefficient.

[0033] For example, when the registration accuracy is high and the drying time is less than 2 seconds, the wetting effect coefficient is close to 0.9, indicating severe solvation penetration between the subsequent ink and the previous wet film. When the registration accuracy is low and the drying time exceeds 10 seconds, the wetting effect coefficient drops below 0.2, indicating weak interaction between ink layers. During the table lookup process, the registration accuracy level serves as a discrete index to directly locate the table row, and the drying time is determined by linear interpolation to determine the column position. The value at the intersection of the two is the required wetting effect coefficient.

[0034] For example, the wetting effect coefficient reflects the diffusion rate of ink molecules at the interface in practical applications. When the coefficient is high, it indicates that the solvent in the previous layer ink has not fully evaporated, and the solvent in the subsequent ink forms a continuous liquid phase channel with the previous layer, promoting abnormal migration of pigment particles.

[0035] S103. Extract spectral features from the chromaticity values ​​of the interface region where the front and rear color layers are in contact and identify hue shifts to obtain hue shift values. Based on the chromaticity value deviation and the dynamic threshold based on the degree of wetting influence, mark the color mixing region and output the region coordinates. Extract the contact area of ​​the front and rear color layers from the region coordinates.

[0036] Reflectance spectral data of the interface between the front and rear color layers is collected within a preset wavelength range of 400 to 700 nanometers. Principal component analysis is performed on the reflection spectral data to extract the first three principal components as spectral feature vectors. The difference between the actual hue angle and the standard hue angle is calculated based on the spectral feature vectors to obtain the hue shift value. Based on the degree of wetting influence and a preset baseline threshold of 0 to 1, a dynamic threshold is calculated by multiplying the values ​​according to empirical values. If the hue shift value exceeds the dynamic threshold, the pixel is determined to be a color mixing point. Consecutive color mixing points are marked with eight-neighbor connected components to form a color mixing region. The coordinates of the four vertices of the minimum bounding rectangle are extracted from the boundary pixel coordinates of the color mixing region. The length and width of the rectangle are calculated using the vertex coordinates. The contact area between the front and rear color layers is obtained by multiplying the length and width. A sliding window is used to traverse the interior of the color mixing region. The rate of change of chromaticity values ​​between adjacent pixels within each window is calculated as the color difference gradient value. If the gradient value shows a decreasing trend from the boundary to the interior, it is determined to be a gradual color mixing caused by solvent diffusion. The set of coordinates of the color mixing region is output.

[0037] Specifically, in one implementation, the spectral module uses a linear CCD sensor combined with a diffraction grating to acquire the spectrum of the printed interface of the trademark tape.

[0038] Specifically, when white light shines on the interface area where the front and rear color layers meet, light of different wavelengths is selectively absorbed and reflected by the ink. The reflected light is then dispersed by a diffraction grating and projected onto a CCD sensor to form spectral data. The preset wavelength range typically covers the visible spectrum. The sampling interval is dynamically adjusted according to the optical properties of the ink; for example, the sampling interval is reduced to 1 nanometer near the absorption peak to improve resolution. Principal component analysis decomposes the spectral data matrix into eigenvalues ​​of the covariance matrix, extracting the eigenvectors corresponding to the three largest eigenvalues ​​as principal components. These three principal components represent the main trend of the spectral curve, secondary fluctuation characteristics, and detailed texture information, respectively. In the calculation of hue shift, the eigenvectors constructed based on principal components are mapped to the CIEXYZ color space through a linear transformation. The input is the eigenvector v, and the output is the XYZ coordinates. The actual hue angle h is calculated using the formula h=arctan(Y / X). This angle is then compared with the theoretical hue angle of the corresponding color layer in the standard printed color spectrum, based on the preset color value library of the Pantone system. The resulting angle deviation is the hue shift value. When magenta and yellow on the trademark band wet and mix, an abnormal bulge appears in the spectral curve near 550 nm at the interface. Principal component analysis can accurately capture this spectral distortion feature, and the calculated hue shift value directly reflects the severity of color mixing.

[0039] It should be noted that the dynamic threshold setting mechanism takes into account the influence of ink wetting state on the color mixing judgment criteria. The baseline threshold is an empirical value obtained through statistical analysis of a large number of samples under standard printing conditions, while the degree of wetting influence is used as an adjustment factor to correct the baseline threshold.

[0040] For example, when the wetting effect is high, solvent migration between ink layers intensifies, and even small hue shifts may indicate serious color mixing problems, at which point the dynamic threshold is correspondingly lowered. The eight-neighbor connected component labeling algorithm starts with a color mixing point and checks whether its eight surrounding pixels are also labeled as color mixing points; if so, they are grouped into the same color mixing region. This labeling method can accurately identify irregularly shaped color mixing regions, including dotted color mixing caused by uneven printing pressure and banded color mixing formed by registration misalignment.

[0041] Preferably, the minimum bounding rectangle is extracted using a rotating caliper algorithm, which finds the rectangle with the smallest area that can completely contain the region by traversing the convex hull vertices of the boundary of the mixed color region.

[0042] In one possible implementation, the long axis of the rectangle is typically parallel or perpendicular to the direction of the label band's movement, reflecting the characteristic that color mixing primarily spreads along the printing direction or laterally. The calculation of the contact area includes not only the geometric area of ​​the rectangle but also the voids and discontinuities within the color mixing region. By counting the actual color-mixing pixels within the rectangle and multiplying this count by the physical area corresponding to each individual pixel, a more accurate value for the actual contact area between the preceding and following color layers is obtained.

[0043] In one embodiment, the size of the sliding window is adaptively adjusted according to the size of the color mixing region, typically set to one-tenth of the minimum side length of the region. The window moves within the color mixing region at preset step sizes, and the average chromaticity value of all pixels within the window is calculated at each position. The color difference gradient value is obtained by calculating the Euclidean distance between the average chromaticity values ​​of adjacent windows and then dividing it by the distance between the center of the windows.

[0044] Specifically, if the color difference gradient value shows a monotonically decreasing trend as it moves from the boundary of the color mixing area inward, it indicates that the color change gradually becomes more gradual, which is a typical characteristic of solvent diffusion. When solvent molecules penetrate from the subsequent ink layer to the previous wet film, a concentration gradient is formed at the interface, which drives the migration of pigment particles, causing the color to gradually transition from deep mixing at the boundary to slight mixing in the interior. Conversely, if the gradient value shows irregular fluctuations or abrupt changes, it may be due to mechanical factors such as printing plate wear or uneven pressure causing color mixing.

[0045] Understandably, the output format of the color mixing region coordinate set includes the boundary coordinate sequence, center coordinates, area size, and color mixing type identifier for each color mixing region.

[0046] For example, in a typical trademark design, when red ink is overprinted on incompletely dried yellow ink, the resulting orange transition band at the interface is accurately identified and marked, and a coordinate set records the precise location and extent of the transition band. Furthermore, by analyzing the distribution pattern of the mixed-color areas across the entire trademark band, the operating status of the printing equipment can be inferred. If the mixed-color areas are mainly concentrated at the edges of the trademark band, it suggests a possible tension imbalance; if the mixed-color areas are periodically distributed, it may be related to the eccentricity of the printing cylinder. In another implementation, the calculation of the color difference gradient value also incorporates directional information, calculating not only the magnitude of the gradient but also recording its direction angle. By analyzing the distribution characteristics of the gradient direction field, the diffusion patterns of the mixed-color areas can be identified, such as radial diffusion, unidirectional diffusion, or vortex diffusion. These patterns correspond to different printing defect mechanisms.

[0047] Specifically, the output set of coordinates for the mixed color areas provides a quantitative basis for subsequent print quality assessment. Each coordinate point is associated with its hue offset value, gradient value, and color mixing type, forming a complete description of the color mixing characteristics, thus enabling accurate identification and location of wetting and dissolving phenomena in multi-color overprinting of trademarks.

[0048] S104. Separate the spectral signals of the front and rear color layers from the corresponding positions of the marked color mixing area, classify and group the spectral signals of the front and rear color layers according to hue, and obtain the color distribution map of each color layer.

[0049] The original spectral data is read from the coordinates of the marked color mixing region. An independent component analysis algorithm is used to perform blind source separation on the mixed spectral signal. The spectral curve characteristics of each independent component are matched with the preset characteristic spectra of cyan, magenta, yellow, and black inks to obtain the separated front and back layer spectral signals. These signals are then converted to the International Commission on Illumination (CIE) L*a*b* color space (CIELab chromaticity space). The hue value of each signal is calculated, and K-means clustering is used to group them according to hue similarity. The input is the hue value dataset of the separated signals, K is set to 4, Euclidean distance is used as the metric, and the convergence condition is that the change in the center value is less than 0.001. The center value and number of members of each hue group are then determined. Based on the spatial distribution density of the members of each hue group, the chromaticity intensity value of each color layer at different locations is statistically analyzed. A continuous chromaticity distribution map is constructed by two-dimensional bilinear interpolation. The input is a discrete grid of chromaticity intensity values, and the output is the interpolated continuous distribution map. The chromaticity distribution map reflects the spatial variation characteristics of the color concentration of each color layer.

[0050] Specifically, in one implementation, the independent component analysis algorithm separates mixed spectral signals by finding statistically independent components.

[0051] Specifically, the algorithm assumes that the spectral signal of the mixed region is a linear superposition of the spectra of the preceding and following ink layers, and recovers the original signal by maximizing the non-Gaussianity of each component. In label printing, cyan ink has an absorption peak at 630 nm, magenta at 550 nm, yellow at 450 nm, and black absorbs uniformly across the entire wavelength range; these characteristic spectra serve as reference templates. During the separation process, the algorithm iteratively adjusts the separation matrix to maximize the correlation coefficient between the output independent components and the reference template. A successful match is determined when the correlation coefficient exceeds 0.85.

[0052] It should be noted that the CIELab color space was chosen based on its perceptual uniformity, where the color difference calculated is linearly related to the color difference perceived by the human eye. The conversion process first integrates the spectral data using a color matching function to obtain tristimulus values, and then maps them to the Lab coordinate system through a nonlinear transformation. The hue value is obtained by calculating the arctangent angles of the a and b components, ranging from 0 to 360 degrees.

[0053] Preferably, the initial cluster centers of the K-means clustering algorithm are selected using the K-means++ method to avoid the local optimum problem caused by random initialization.

[0054] In one possible implementation, the number of clusters K is dynamically determined based on the number of design colors in the trademark pattern. This number refers to the number of unique colors in the pattern, which is obtained statistically through image analysis software and is typically between 4 and 8. The input to K-means clustering is the pixel hue value of the mixed-color area (i.e., the part where colors mix) in the trademark pattern, and the output is the hue group. The number of members in each group reflects the proportion of that color in the mixed-color area, and the center value represents the dominant hue of that group.

[0055] For example, when red ink and yellow ink wet and dissolve together, the clustering results will show an orange group, whose number of members and their distribution location directly indicate the range and degree of color mixing.

[0056] For example, two-dimensional interpolation employs a bicubic interpolation method to construct a smooth chromaticity variation surface between discrete sampling points.

[0057] Specifically, for each spatial location, the interpolation result is calculated using a cubic polynomial function based on the chromaticity values ​​of 16 known points around it. The chromaticity distribution map is displayed in pseudo-color, with different color intensities represented by gradient levels, from dark blue representing low chromaticity to dark red representing high chromaticity. Furthermore, the chromaticity distribution map can visually reflect the spatial diffusion pattern of ink. When the distribution map exhibits a bell-shaped distribution with a high center and low edges, it indicates that the ink concentration decreases from the center of the printing area outwards; when multiple peaks appear, it suggests the presence of local ink accumulation or printing plate defects. By analyzing the superposition relationship of the chromaticity distribution maps of each color layer, the interaction strength and color mixing mechanism between ink layers can be determined.

[0058] S105. The hue shift value is comprehensively evaluated with the drying time and contact area of ​​the previous layer ink. The degree of color mixing at the interface is extracted by combining the color distribution map of each color layer. The degree of color mixing is compared with the adjustment threshold based on the overprinting accuracy level to identify abnormal color mixing caused by solvent wetting and output an indicator of increased color difference.

[0059] A three-dimensional evaluation matrix is ​​constructed based on hue shift values, previous ink drying time, and contact area. Normalization is applied to ensure all dimensions are within a unified dimension. The three dimensions are then weighted and summed using preset weighting coefficients to obtain a comprehensive evaluation value. Boundary gradient peaks and central chromaticity concentrations are extracted from the chromaticity distribution maps of each color layer. The ratio of the boundary gradient peak to the central chromaticity concentration is calculated. If this ratio exceeds a preset interface threshold, interface mixing is identified. The percentage of pixels involved in interface mixing is statistically analyzed to determine the degree of interface mixing. Based on printing accuracy levels, the classification criteria are: high accuracy deviation less than 0.1mm, medium accuracy 0.1-0.5mm, and low accuracy greater than 0.5mm. This is determined using printing deviation measurement methods. The corresponding adjustment threshold is retrieved from a preset threshold table. If the product of the interface mixing degree and the comprehensive evaluation value exceeds the adjustment threshold, it is determined to be abnormal mixing caused by solvent wetting. An increased color difference indicator is assigned to the abnormal mixing area. The location coordinates, mixing degree value, and trigger threshold of the abnormal mixing are recorded. Detection result data including the increased color difference indicator is output.

[0060] Specifically, in one implementation, the process of constructing the three-dimensional evaluation matrix involves unifying parameters of three different physical dimensions.

[0061] Specifically, hue shift values ​​are measured in degrees, typically ranging from 0 to 180 degrees; the drying time of the previous ink layer is measured in seconds, usually 0.5 to 30 seconds in high-speed printing; and the contact area is measured in square millimeters, varying considerably depending on the complexity of the trademark design. Normalization employs a min-max standardization method, mapping each parameter to a uniform dimension range of 0 to 1. Weighting coefficients are set based on statistical analysis of a large number of printed samples. The weight for hue shift values ​​is typically set to 0.5, as it directly reflects the degree of color change; the drying time weight is set to 0.3, reflecting the influence of ink wetness; and the contact area weight is set to 0.2, characterizing the spatial range of color mixing. The comprehensive evaluation value obtained through weighted summation fully reflects the combined impact of multiple factors on color mixing. When the comprehensive evaluation value approaches 1, it indicates a serious risk of color mixing.

[0062] It should be noted that the extraction of boundary gradient peaks and center chromaticity concentration is based on spatial feature analysis of the chromaticity distribution map. The boundary gradient peak is obtained by calculating the rate of chromaticity change between adjacent pixels at the edge of the chromaticity distribution map; a higher peak indicates a sharper boundary of the color layer. The center chromaticity concentration is obtained by averaging the chromaticity values ​​of all pixels within the central region of the color layer, reflecting the overall ink concentration level.

[0063] Preferably, the physical meaning of the ratio lies in quantifying the degree of ink diffusion from the boundary to the center. When the ratio is close to 1, it indicates that the color of the boundary and the center are basically the same, and the ink distribution is uniform. When the ratio is greater than a preset interface threshold, it indicates that there is abnormal color aggregation or diffusion at the boundary, which is a typical characteristic of interface color mixing. The percentage of pixels with interface color mixing is obtained by dividing the number of pixels exceeding the threshold by the total number of pixels. This value directly reflects the extent of the color mixing effect.

[0064] In one possible implementation, the preset threshold table is a two-dimensional lookup table established through experimental calibration. The horizontal axis represents the registration accuracy level, including high, medium, and low levels; the vertical axis represents different combinations of printing conditions, such as ink type, substrate, and printing speed. Each cell stores the adjustment threshold for the corresponding condition.

[0065] For example, when using high-precision overprinting with water-based inks on coated paper, the threshold is set to 0.6; when using low-precision overprinting with solvent-based inks on thin film materials, the threshold is relaxed to 0.8. The query process directly locates the corresponding threshold value through a combination index of precision level and printing conditions.

[0066] For example, the product of the interface color mixing degree and the overall evaluation value forms a comprehensive judgment index. This product calculation takes into account the interaction between local color mixing characteristics and the overall evaluation result. When the interface color mixing degree is 0.7 and the overall evaluation value is 0.8, the product is 0.56. If this value exceeds the adjustment threshold of 0.5, it is judged as abnormal color mixing caused by solvent wetting. This judgment method can distinguish between registration deviation caused by mechanical factors and wetting miscibility caused by chemical factors.

[0067] Specifically, the color difference increase indicator uses a structured data format to record the detection results. Each indicator contains five fields: a region number to uniquely identify the mixed area; location coordinates to record the center point and boundary range of the mixed area; a color mixing degree value to quantify the severity of the color mixing; a trigger threshold to record the threshold parameter used in the judgment; and a timestamp to mark the time when the detection occurred. Furthermore, the output format of the detection result data supports multiple application scenarios. In real-time monitoring mode, the data is streamed to the printing control system to trigger the automatic adjustment mechanism; in offline analysis mode, the data is stored as a structured file for quality analysts to perform statistical analysis and process improvement.

[0068] Understandably, the identification results of abnormal color mixing can guide the optimization and adjustment of the printing process. When a batch of trademark tapes shows a large number of increased color difference indicators, the root cause of the problem can be located by analyzing the spatial distribution pattern and temporal evolution trend of the indicators. If the indicators are concentrated at the edge of the trademark tape, it indicates a problem with tension control; if the indicators are periodically distributed, it may be related to the mechanical vibration of the printing cylinder. In another embodiment, the increased color difference indicators also include suggested adjustment parameters, such as increasing drying time, reducing printing speed, or adjusting ink viscosity. These parameters are automatically recommended based on a rule base established from historical data and expert experience, providing operators with direct improvement guidance and realizing closed-loop control from detection to correction.

[0069] S106. Match and classify the color difference increase markers and overprint offsets in the same area according to the area location to generate an overprint quality assessment report. Optimize the recognition threshold based on the abnormal distribution in the assessment report to obtain an updated overprint quality recognition benchmark. Use the updated benchmark to batch process subsequent images to extract the correspondence between the offset and the degree of color mixing. If the overall color difference change meets the range of color mixing, output the final overprint quality assessment result.

[0070] Spatial matching is performed based on the location coordinates of the color difference increase markers and the region code of the overprint offset. A hash table is used to establish an index mapping from coordinates to regions. By querying the index, the markers and offsets within the same region are categorized and aggregated. A data structure containing region number, marker quantity, average offset, and color mixing type is constructed to obtain an overprint quality assessment report. The density of abnormal markers in each region is statistically analyzed from the overprint quality assessment report, and the mean and standard deviation of the abnormal density are calculated. If the abnormal density of a region exceeds the mean plus a preset multiple of the standard deviation, it is marked as a high-incidence region. The recognition threshold is adjusted based on the difference ratio between the offset feature value of the high-incidence region and the normal region. The adjusted recognition threshold is used as the update benchmark to process subsequent batches of images, extracting the overprint offset and interface color mixing degree of each image, establishing a corresponding dataset of offset and color mixing degree, and determining the correspondence between the two through linear regression. Based on the correspondence, the predicted value of the overall color difference change is calculated. If the deviation between the predicted value and the actual measured color difference value is within a preset tolerance range, the color difference change is determined to meet the color mixing degree range, and the final overprint quality assessment result is output.

[0071] Specifically, in one implementation, the hash table is constructed using a spatial grid partitioning method to achieve a fast mapping from location coordinates to region codes.

[0072] Specifically, the printing area of ​​the trademark strip is divided into several rectangular regions according to a preset grid size, and each region is assigned a unique code identifier. The position coordinates of the anomaly identifier are directly calculated by dividing by the grid size and rounding down to obtain the corresponding grid index, which is stored as a hash key in a hash table. The overprinting offset data obtained from the printing inspection equipment is assigned to a region according to the same grid division rules. By leveraging the constant time complexity query characteristic of the hash table, rapid matching of a large number of identifiers and offset data is achieved. During the classification and aggregation process, all identifiers and offsets with the same region code are collected into the same data structure. This structure includes the region number as the primary key, the cumulative value of the number of identifiers, the arithmetic mean of all offsets, and the color mixing type, which refers to the color difference category caused by the superposition of different colors, such as the classification statistics of the mixing of cyan, magenta, yellow, and black. The overprinting quality assessment report is output in the form of a structured table, with each row corresponding to a region and the columns including the above statistical indicators, forming a panoramic view of the quality status at the region level.

[0073] It should be noted that the method for calculating the density of anomalous markers directly affects the identification accuracy of high-incidence areas. Anomaly density is defined as the number of anomalous markers per unit area, obtained by dividing the total number of markers in the area by the area of ​​the region.

[0074] Preferably, the statistical analysis of the mean and standard deviation is based on density data from all regions. The standard deviation reflects the dispersion of the density distribution. When the abnormal density of a region exceeds the mean plus a preset multiple of the standard deviation, it indicates that the abnormality of that region significantly deviates from the normal level. The preset multiple is usually 2 or 3, corresponding to the 95% or 99% confidence interval in statistics. The offset characteristic values ​​of high-incidence regions include the consistency of the offset direction, the concentration of the offset amplitude, and the regularity of the offset pattern. The identification threshold is adjusted using a scaling method. If the average offset of a high-incidence region is 1.5 times that of a normal region, the identification threshold for that region is reduced to two-thirds of the original value, so that the detection sensitivity matches the actual color mixing risk.

[0075] In one possible implementation, the updated recognition benchmark forms a region-specific detection standard system. Batch processing employs a pipelined parallel processing architecture, with multiple images simultaneously entering different stages of the processing pipeline, including image reading, feature extraction, offset calculation, and color mixing degree evaluation. Each image generates a set of offset-color mixing degree data pairs, which constitute the sample set for regression analysis. Linear regression fits the linear relationship y=ax+b between offset x and color mixing degree y using the least squares method, where the slope a reflects the strength of the offset's influence on the color mixing degree, and the intercept b represents the baseline color mixing level at zero offset.

[0076] For example, the establishment of the correspondence reveals the intrinsic relationship between printing accuracy and color mixing. When the coefficient of determination R² of the linear regression is greater than 0.8, it indicates that the offset can explain the variation in the degree of color mixing well, verifying the dominant role of the wetting and mixing mechanism.

[0077] Specifically, the predicted value of the overall color difference change is calculated by substituting the offset of the new image into the regression equation. The actual measured color difference value is obtained using a spectrophotometer under a standard light source and is expressed as a CIE ΔE value. The preset tolerance range is set according to printing industry standards, typically ±2.0 of the ΔE value. When the deviation between the predicted and actual values ​​falls within the tolerance range, it indicates that the established correspondence model is accurate and reliable, capable of predicting the final color difference performance through offset and color mixing degree. Furthermore, the final overprint quality assessment result includes multi-dimensional quality indicators. In addition to the overall pass / fail judgment, it also provides a quality level distribution map for each region, spatial location markers of areas with high incidence of anomalies, the correlation coefficient between offset and color mixing degree, and suggested process adjustment directions. In another embodiment, the assessment result also includes a time series analysis function. By performing time series analysis on the assessment results of consecutive batches, it identifies patterns of quality trend changes, such as periodic fluctuations, monotonous deterioration, or abrupt anomalies. This dynamic monitoring mechanism can provide early warning of potential equipment failures or process malfunctions, enabling preventative maintenance.

[0078] Based on the embodiments of the present invention described above, and through the above description, those skilled in the art can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for automatic identification and evaluation of color difference in trademark printing, characterized in that, include: Extract the boundary positions of each color layer from the captured multi-color overprinted image of the trademark, compare the boundary positions of each color layer with the standard overprinting position, and obtain the overprinting offset of each region. The overprinting accuracy level is determined based on the overprinting offset of each region. The drying time of the previous layer ink is extracted from the printing process record. The overprinting accuracy level is matched with the drying time of the previous layer ink to obtain the degree of wetting effect. Spectral features are extracted from the chromaticity values ​​of the interface region between the front and rear color layers to obtain hue shift values. The hue shift values ​​are compared with a dynamic threshold based on the degree of wetting influence to mark the color mixing region and output the region coordinates. The contact area between the front and rear color layers is extracted from the region coordinates. The spectral signals of the front and rear color layers are separated from the corresponding positions in the color mixing region, and grouped according to hue to obtain the color distribution map of each color layer; A comprehensive evaluation is performed based on the color distribution diagrams of each color layer to obtain the degree of color mixing at the interface. The degree of color mixing at the interface is compared with the adjustment threshold based on the overprinting accuracy level, and an indicator of increased color difference is output. The color difference increase indicator and the overprint offset are matched and categorized according to their regional positions to generate an overprint quality assessment report.

2. The method for automatic identification and evaluation of color difference in trademark printing according to claim 1, characterized in that, The step of extracting the boundary positions of each color layer from the captured multi-color overprinted image of the trademark, comparing the boundary positions of each color layer with the standard overprinting position, and obtaining the overprinting offset of each region includes: performing edge gradient detection on the multi-color overprinted image of the trademark using the Sobel operator to obtain the edge gradient intensity value of each color layer; constructing a boundary contour set based on pixels whose gradient intensity values ​​exceed a preset threshold; filtering closed contours from the boundary contour set and calculating the centroid coordinates of the contours; calculating the Euclidean distance between the centroid coordinates of the contours and the standard overprinting reference coordinates to determine the offset anomaly points; performing spatial clustering on the offset anomaly points using a density clustering method; dividing the overprinting region and generating a region code based on the clustering results; extracting the corner points at the color layer intersections from the image blocks corresponding to the region codes as alignment feature points; fitting the offset vector between the alignment feature points and the standard position using the least squares method; calculating the components of the offset vector in the horizontal and vertical directions to obtain the overprinting offset of each region.

3. The method for automatic identification and evaluation of color difference in trademark printing as described in claim 1, characterized in that, The process of determining the registration accuracy level based on the registration offset of each region, extracting the drying time of the previous layer ink from the printing process record, and matching the registration accuracy level with the drying time of the previous layer ink to obtain the degree of wetting influence includes: comparing the value of the registration offset of each region with the grading threshold to determine a high accuracy level, medium accuracy level, or low accuracy level; reading the completion time stamp of the previous layer ink printing and the start time stamp of the subsequent layer ink printing from the printing equipment process record database, calculating the difference between the completion time stamp of the previous layer ink printing and the start time stamp of the subsequent layer ink printing to obtain the drying time of the previous layer ink; and using the registration accuracy level and the drying time of the previous layer ink as an index to look up the wetting influence mapping table to obtain the wetting influence coefficient as the degree of wetting influence.

4. The method for automatic identification and evaluation of color difference in trademark printing according to claim 1, characterized in that, The process of extracting spectral features from the chromaticity values ​​of the interface region between the front and rear color layers to obtain a hue shift value, comparing the hue shift value with a dynamic threshold based on the degree of wetting influence, marking the mixed color region and outputting the region coordinates, and extracting the contact area between the front and rear color layers from the region coordinates includes: collecting reflectance spectral data of the interface between the front and rear color layers; performing principal component analysis on the reflectance spectral data to extract the first three principal components as spectral feature vectors; calculating the difference between the actual hue angle and the standard hue angle to obtain the hue shift value; calculating a dynamic threshold based on the degree of wetting influence; marking pixels whose hue shift value exceeds the dynamic threshold with eight-neighbor connected components to form a mixed color region and outputting a set of mixed color region coordinates; extracting the vertex coordinates of the minimum bounding rectangle from the boundary pixel coordinates of the mixed color region; calculating the product of the rectangle's length and width to obtain the contact area between the front and rear color layers.

5. The method for automatic identification and evaluation of color difference in trademark printing according to claim 1, characterized in that, The step of separating the spectral signals of the front and rear color layers from the corresponding positions of the color mixing region, grouping them according to hue, and obtaining the chromaticity distribution map of each color layer includes: using an independent component analysis algorithm to perform blind source separation on the original spectral data of the color mixing region; matching the spectral curve after blind source separation with the characteristic spectra of the four-color inks to obtain the spectral signals of the front and rear color layers; converting the spectral signals of the front and rear color layers to the CIE Lab chromaticity space, calculating the hue value, and grouping them according to hue similarity using K-means clustering; statistically analyzing the chromaticity intensity values ​​based on the spatial distribution density of each hue group member, and constructing the chromaticity distribution map of each color layer through two-dimensional interpolation.

6. The method for automatic identification and evaluation of color difference in trademark printing according to claim 1, characterized in that, The process involves comprehensively evaluating the interface color mixing degree based on the chromaticity distribution maps of each color layer, comparing the interface color mixing degree with an adjustment threshold based on the overprinting accuracy level, and outputting a color difference increase indicator. This includes: normalizing the hue shift value, the drying time of the previous layer ink, and the contact area, and then performing a weighted summation to obtain a comprehensive evaluation value; extracting the boundary gradient peak and the center chromaticity concentration from the chromaticity distribution maps of each color layer, calculating the ratio of the boundary gradient peak to the center chromaticity concentration, and statistically analyzing the pixel proportions to obtain the interface color mixing degree; querying the adjustment threshold based on the overprinting accuracy level, and if the product of the interface color mixing degree and the comprehensive evaluation value exceeds the adjustment threshold, then assigning a color difference increase indicator to the abnormal area and recording the location coordinates and color mixing degree value.

7. The method for automatic identification and evaluation of color difference in trademark printing according to claim 1, characterized in that, The step of matching and classifying the color difference increase markers and the overprint offsets according to their regional locations to generate an overprint quality assessment report includes: performing spatial matching based on the position coordinates of the color difference increase markers and the regional codes of the overprint offsets, establishing a hash table index, realizing the classification and aggregation of markers and offsets within the same region, and generating a data structure containing region number, number of markers, average offset, and color mixing type as the overprint quality assessment report.

8. The method for automatic identification and evaluation of color difference in trademark printing according to claim 1, characterized in that, After generating the overprint quality assessment report, the recognition threshold is optimized and the overprint quality recognition benchmark is updated based on the assessment report. The updated benchmark is then used to batch process subsequent images, and the final overprint quality assessment result is output.

9. The method for automatic identification and evaluation of printing color difference in trademarks according to claim 8, characterized in that, The process of optimizing the identification threshold and updating the overprint quality identification benchmark based on the evaluation report, and using the updated benchmark to perform batch processing on subsequent images to output the final overprint quality evaluation result includes: statistically analyzing the density of abnormal markers in the overprint quality evaluation report, adjusting the identification threshold based on the offset characteristics of high-incidence areas to obtain the updated benchmark; using the updated benchmark to extract the overprint offset and interface color mixing degree from subsequent batch images, establishing the correspondence between the overprint offset and the interface color mixing degree, and outputting the final overprint quality evaluation result if the overall color difference variation deviation is within the tolerance range.

Images