Visual prediction method and system for color register deviation of multi-color printing machine

Abstract

A visual prediction method and system for color register deviation of a multicolor press belong to the technical field of printing image detection, and the method comprises the following steps: obtaining a first reflection image of a printing material collected under the irradiation of a first illumination mode; according to the first reflection image, a registration mark is positioned, and a color register deviation value is calculated; if the color register deviation value exceeds the preset deviation threshold value, evaluating the confidence coefficient of the color register deviation value according to the first reflection image; when the confidence coefficient is lower than a preset confidence coefficient threshold value, acquiring a second reflection image of the printing material collected under the irradiation of a second illumination mode; and according to the second reflection image, judging whether the suspicious abnormal state is caused by surface attachment interference of the printing material. According to the scheme, not only can the color register deviation be detected, but also the potential reason causing the deviation can be detected, so that a large number of false alarms caused by interference of surface attachments are avoided, the non-planned downtime is shortened, and the production efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of printing image detection technology, specifically to a visual prediction method and system for color registration deviation in multi-color printing presses. Background Technology

[0002] In the precision printing, packaging, and high-end materials processing industries, the printing accuracy of product patterns is one of the core quality indicators. To ensure accurate registration of multi-color patterns, the industry commonly uses online inspection systems based on machine vision. These systems capture printing marks in real time using cameras, calculate the relative positional deviation between each color plate (i.e., registration deviation), and provide feedback to control the printing unit for dynamic adjustment. With the continuous improvement of production speed and quality requirements, increasingly stringent requirements have been placed on the speed, accuracy, and especially the reliability of registration deviation detection. Any unplanned downtime caused by false alarms can result in significant economic losses.

[0003] Currently, mainstream color misregistration visual inspection systems primarily rely on reflected images under a single, fixed illumination method (such as low-angle ring light, coaxial light, or diffused light). The system optimizes these illumination conditions to achieve stable contrast at the printed markings, enabling the algorithm to perform sub-pixel-level precise positioning and misregistration calculation. This method is relatively effective when inspection conditions are ideal and the substrate surface is stable. Some systems, in order to cope with complex scenarios, attempt to integrate multiple light sources or employ more complex image processing algorithms, but their core decisions are still based on two-dimensional image information obtained under a single physical imaging principle.

[0004] However, in real-world high-speed production environments, the complex surface conditions of printing materials affect the reliability of the aforementioned single-mode detection methods. In particular, random foreign matter such as dust, oil, and water stains adhering to the material surface can create local image features highly similar to registration misalignment under specific lighting conditions, leading to numerous false alarms. Simply relying on relaxing the threshold to reduce false alarms increases the risk of missed detections. Therefore, the industry urgently needs a registration misalignment detection method capable of diagnosing suspected deviations to fundamentally solve the false alarm problem. Summary of the Invention

[0005] This application provides a visual prediction method, apparatus, equipment, and storage medium for color registration deviation in multicolor printing presses, which can solve the technical problems existing in the above-mentioned related technologies.

[0006] In a first aspect, embodiments of this application provide a visual prediction method for color registration deviation in a multi-color printing press, employing the following technical solution:

[0007] A visual prediction method for color registration deviation in a multi-color printing press includes the following steps:

[0008] Acquire a first reflected image of the printed material under a first illumination mode; wherein, under the first illumination mode, the light source provides directional illumination to the surface of the printed material to highlight the geometric edges of the registration marks on the printed material;

[0009] Based on the first reflection image, locate the registration mark and calculate the color deviation value;

[0010] If the color mismatch value exceeds a preset mismatch threshold, the confidence level of the color mismatch value is evaluated based on the first reflection image;

[0011] When the confidence level is lower than the preset confidence level threshold, it is determined to be a suspicious abnormal state;

[0012] In response to the suspected abnormal state, a second reflection image of the printed material under the illumination of the second lighting mode is acquired; wherein, under the second lighting mode, the light source provides uniform diffuse illumination to the surface of the printed material to highlight the surface texture of the printed material.

[0013] Based on the second reflection image, determine whether the suspected abnormal state is caused by interference from substances adhering to the surface of the printed material;

[0014] If so, output a prompt signal indicating that the color mismatch value is caused by surface interference of the printing material.

[0015] In conjunction with the first aspect, in one embodiment, assessing the confidence level of the color misalignment value based on the first reflective image includes the following steps:

[0016] Extract various visual evaluation features related to the registration mark from the first reflection image;

[0017] A normalized score is applied to each of the aforementioned visual evaluation features;

[0018] The normalized scores are combined using a weighted geometric mean algorithm to obtain a confidence level within a set range.

[0019] In conjunction with the first aspect, in one implementation, the plurality of visual evaluation features includes at least two of edge sharpness features, template matching confidence features, region contrast features, and location residual features;

[0020] Wherein, the edge sharpness feature is a characterization value reflecting the clarity of the registration mark edge;

[0021] The template matching confidence feature is a score value calculated by combining multiple related peak morphology information in the related peaks generated after matching the preset registration mark template with the first reflection image; the multiple related peak morphology information includes at least two of the following: the sharpness of the related peak, the signal-to-noise ratio of the related peak, and the intensity ratio of the main peak to the secondary peak;

[0022] The region contrast feature is a characterization value that reflects the degree of grayscale difference between the region where the registration mark is located and the background region;

[0023] The position residual feature is a characterization value reflecting the deviation between the position of the registration mark obtained in this calculation and the expected position predicted based on historical position data.

[0024] In conjunction with the first aspect, in one embodiment, determining whether the suspected abnormal state is caused by interference from deposits on the surface of the printed material based on the second reflection image includes the following steps:

[0025] The analysis region in the second reflection image is determined based on the position of the registration mark in the first reflection image;

[0026] Calculate at least two discriminant features of the analysis region; the discriminant features are selected from the pixel window unit texture contrast of the analysis region, the shape irregularity of the gray-level binary region in the analysis region, and the structural similarity between the analysis region and the corresponding region in the preset historical cleaning reference image of printing materials;

[0027] Based on at least two of the discriminative features, it is determined whether the suspected abnormal state is caused by interference from surface deposits.

[0028] In conjunction with the first aspect, in one implementation, after determining a suspicious abnormal state when the confidence level is lower than a preset confidence level threshold, the method further includes the following steps:

[0029] At least two third reflection images of the printed material are acquired under the illumination of a third illumination mode; wherein the third illumination mode is a linearly polarized light source that provides polarized light illumination to the surface of the printed material in the same direction as the first illumination mode, the third reflection images are captured by a camera combined with a polarization state analyzer, and the two third reflection images have different polarization state configurations.

[0030] Based on the two third reflection images, polarization feature information related to the optical anisotropy of the material is calculated; the polarization feature information is a polarization degree image or a polarization difference image.

[0031] Based on the polarization characteristic information, determine whether the suspected abnormal state is caused by internal stress or deformation of the printing material;

[0032] If so, output a warning signal indicating that the color mismatch value is caused by internal stress or deformation of the printing material.

[0033] In conjunction with the first aspect, in one embodiment, the two third reflection images have different polarization state configurations.

[0034] The transmission axis of the polarization analyzer corresponding to the two third reflection images forms an angle of 0 degrees and 90 degrees with the polarization direction of the linearly polarized light source, respectively.

[0035] In conjunction with the first aspect, in one embodiment, determining whether the suspected abnormal state is caused by internal stress or deformation of the printing material based on the polarization feature information includes the following steps:

[0036] Analyze whether the polarization characteristic information exhibits regular interference fringes or significant birefringence characteristics within the region where the registration mark is located;

[0037] If so, the suspected abnormal state is determined to be caused by internal stress or deformation of the material;

[0038] If not, then the suspected abnormal state is determined not to be caused by internal stress or deformation of the material.

[0039] In conjunction with the first aspect, in one embodiment, when the confidence level is lower than a preset confidence level threshold, the second reflection image is acquired within a set duration threshold, or the second reflection image and the third reflection image are acquired continuously; the duration threshold is calculated based on the conveying speed of the printing material and the length of the area where the registration mark is located in the conveying direction of the printing material.

[0040] In conjunction with the first aspect, in one implementation method.

[0041] Secondly, embodiments of this application provide a visual prediction device for color registration deviation in a multi-color printing press, employing the following technical solution:

[0042] A visual prediction device for color registration deviation in a multi-color printing press, comprising:

[0043] An image acquisition module is configured to acquire a first reflection image of the printed material under illumination in a first lighting mode, and in response to the suspected abnormal state, acquire a second reflection image of the printed material under illumination in a second lighting mode; wherein, in the first lighting mode, the light source provides directional illumination to the surface of the printed material to highlight the geometric edges of the registration marks on the printed material; in the second lighting mode, the light source provides uniform diffuse illumination to the surface of the printed material to highlight the surface texture of the printed material.

[0044] A color misalignment calculation module is configured to locate the registration mark and calculate the color misalignment value based on the first reflection image;

[0045] The color registration deviation result judgment module is configured to: if the color registration deviation value exceeds a preset deviation threshold, evaluate the confidence level of the color registration deviation value based on the first reflection image; if the confidence level is lower than the preset confidence threshold, determine it as a suspicious abnormal state; determine whether the suspicious abnormal state is caused by interference from surface deposits of the printing material based on the second reflection image; if so, output a prompt signal indicating that the color registration deviation value is caused by surface interference of the printing material.

[0046] Thirdly, embodiments of this application provide a visual prediction system for color registration deviation in multi-color printing presses, employing the following technical solution:

[0047] A visual prediction system for color registration deviation in a multi-color printing press, comprising:

[0048] An illumination device includes a light source and a PDLC transmissive plate disposed in the light path between the light source and the printed material. The PDLC transmissive plate can switch between a transparent state and a frosted state. The illumination device provides the printed material with a light mode including at least a first illumination mode and a second illumination mode through the PDLC.

[0049] An imaging device for capturing a corresponding reflected image of the printed material illuminated by the lighting device;

[0050] The visual prediction device for color registration deviation in multi-color printing presses as described above.

[0051] The beneficial effects of the technical solutions provided in this application include:

[0052] This application provides a visual prediction method, apparatus, and system for color registration deviation in multi-color printing presses. Firstly, through directional illumination under a first lighting mode, the system efficiently performs preliminary registration mark positioning and deviation calculation. When a deviation exceeds a threshold, instead of immediately triggering an alarm or adjustment, the system further assesses the confidence level of the deviation value, allowing the system to self-check the reliability of the preliminary detection results. When the confidence level is insufficient to support the current deviation result, the system intelligently triggers diffuse illumination under a second lighting mode and acquires a second reflection image. By analyzing the surface texture features in the second reflection image, the system can acquire more targeted information, thereby determining whether suspicious abnormal states are caused by interference from surface deposits. Compared to traditional systems that cannot distinguish true deviations, this application can not only detect color registration deviations but also detect the potential causes of the deviations, thus avoiding numerous false alarms caused by surface deposit interference in traditional methods. This allows operators to perform accurate problem troubleshooting and handling, reducing unplanned downtime and improving production efficiency. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating an embodiment of the visual prediction method for color registration deviation in a multi-color printing press according to this application.

[0054] Figure 2 This is a schematic diagram of the functional modules of a visual prediction system for color registration deviation in a multi-color printing press according to an embodiment of this application. Detailed Implementation

[0055] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0056] In existing color misregistration visual inspection systems, random foreign objects on the surface of the printing material produce local image features similar to registration misregistration under conventional directional illumination modes. For example, surface deposits such as dust, oil stains, or water stains may form false edge features after illumination and be misjudged as edge features of registration marks, leading to subsequent misjudgments of the registration standard. This results in false alarms and related downtime adjustments, significantly reducing the reliability of the inspection system, increasing the risk of unplanned downtime, and ultimately causing a decrease in production efficiency and affecting the overall system stability.

[0057] To address this issue, this application provides a visual prediction method and system for color registration deviation in multicolor printing presses.

[0058] Firstly, referring to Figure 1 This application provides a visual prediction method for color registration deviation in multi-color printing presses, the method comprising the following steps:

[0059] S100. Acquire a first reflected image of the printing material under the illumination of a first illumination mode; wherein, under the first illumination mode, the light source provides directional illumination to the surface of the printing material to highlight the geometric edges of the registration marks on the printing material.

[0060] Specifically, the first illumination mode is used to form an illumination beam at a certain angle on the surface of the printed material, thereby enabling the edges of the illuminated registration mark to produce obvious light and shadow contrast, so as to form a clearer geometric outline in the first reflected image. For example, a small incident angle can be set between the light source and the surface of the printed material to enhance the shadow effect of the edges, laying the foundation for subsequent accurate identification.

[0061] S200. Based on the first reflected image, locate the registration mark and calculate the color deviation value;

[0062] Specifically, the purpose of step S200 is to calculate the registration deviation value of the registration mark based on the visual information in the first reflection image. This process employs a grayscale threshold segmentation method to identify the edge of the registration mark based on the grayscale value and separate the registration mark region in the first reflection image from the background. Subsequently, by calculating the coordinates of the geometric center or specific feature points of the segmented registration mark and comparing them with the preset ideal position, the relative positional offset between the two, i.e., the registration deviation value, is obtained.

[0063] S300. If the color registration deviation value exceeds the preset deviation threshold, evaluate the confidence level of the color registration deviation value based on the first reflection image.

[0064] Specifically, step S300 is to further evaluate whether the calculated color misalignment value is sufficiently realistic and reliable based on visual information within the first reflection image when the calculated color misalignment value is abnormal. For example, as a possible evaluation method, the average gradient value or local contrast of the registration mark region can be analyzed. For instance, if the edges of the registration mark are blurry, or the contrast of its surrounding area is low, it may indicate that the image quality itself is poor or that there is significant interference. In this case, the confidence level of the color misalignment value obtained from the image can be determined to be low. Conversely, if the edges are sharp and the contrast is high, the confidence level is high.

[0065] S400. When the confidence level is lower than the preset confidence level threshold, it is determined to be a suspicious abnormal state.

[0066] When the confidence level is lower than a preset confidence threshold, it is judged as a suspicious abnormal state. This confidence threshold is set based on historical data obtained from the analysis of possible attachments in the production environment or by experts using their experience in identifying the edge characteristics of possible attachments. For example, when the scores for edge clarity of the registration mark and contrast with the surrounding area displayed in the confidence level are lower than the corresponding confidence threshold, the system marks the current color deviation result as suspicious, indicating that the deviation value may not be caused by actual color misalignment, but rather by other factors.

[0067] S510. In response to a suspected abnormal state, acquire a second reflection image of the printed material under the illumination of a second lighting mode; wherein, in the second lighting mode, the light source provides uniform diffuse illumination to the surface of the printed material to highlight the surface texture of the printed material.

[0068] The second illumination mode differs from the directional illumination of the first illumination mode in that the light source provides uniform diffuse illumination to the surface of the printed material. This reduces shadows and specular reflections on the printed material, thereby more clearly revealing the texture details of the surface, such as attachments and scratches. For example, as one implementation, a ring-shaped diffuse light source or a diffuser integrated into the optical path can be used to provide uniform diffuse illumination to the surface of the printed material. Thus, the image acquisition device can acquire a second reflection image that reflects the surface details.

[0069] S520. Based on the second reflection image, determine whether the suspicious abnormal state is caused by interference from the surface of the printed material.

[0070] Subsequently, based on the second reflection image, it is determined whether the suspected abnormal state is caused by interference from surface deposits on the printed material. Specifically, texture analysis can be performed on the region related to the registration marks in the second reflection image. For example, in some embodiments, the local variance or entropy value of the region can be calculated to characterize the complexity of its texture. If the texture features of the region differ significantly from those of a clean printed material surface, such as the presence of abnormal high-frequency textures or irregular spots, it can be preliminarily determined that surface deposit interference exists.

[0071] S530, If so, output a color registration deviation value as a prompt signal caused by surface interference of the printing material.

[0072] Once the presence of deposits on the printing material surface is confirmed, a prompt signal can be output to inform the operator of the true cause of the abnormal color registration deviation value, thereby avoiding unnecessary production stoppages or adjustments and ensuring production efficiency and system stability. This prompt signal can be presented in various forms, such as displaying the text "Surface Deposits Interference" on the operating interface, or alerting the operator via an audible and visual alarm.

[0073] In the above scheme, after calculating the color registration deviation value, the result is not immediately determined, but a "credibility assessment" step is added. This is because if there are random adhering objects such as dust or oil on the surface of the printing material, these adhering objects may form local shadows, bright spots, or texture changes in the image under directional lighting used for precise positioning. This can result in blurred edges, irregular shapes, or unreasonable abrupt changes in the position of the registration marks, all of which affect the edge features in the first reflection image. Based on this phenomenon, this scheme analyzes the quality of the first reflection image on which the deviation value was generated. If the registration marks in the image have clear edges, regular contours, or a high degree of matching with the standard template, and their positional changes conform to the law of continuous movement, then the calculated deviation value is considered to have high credibility, and it is likely that there are no adhering objects, thus reflecting a true color registration anomaly. Conversely, if the marks in the image have blurred edges, irregular shapes, or unreasonable abrupt changes in position, it means that the current image quality is likely to be interfered with, and the calculated deviation value may include the edge features of related adhering objects as a reference, leading to erroneous calculation results and low credibility. Therefore, the confidence level calculated in this scheme is an internal measure of the reliability of the current deviation value calculation result. When the system detects that the deviation value exceeds the limit, it will first check the confidence level of the result. Only when both the "deviation value exceeds the limit" and "confidence level is too low" conditions are met will the system suspect that the current anomaly may be an artifact caused by surface deposits, thereby triggering subsequent verification using a diffuse illumination mode specifically designed for surface texture. This design enables the system to effectively distinguish between "genuine overprinting errors" and "detection artifacts caused by surface contamination," thereby greatly reducing false alarms caused by random surface interference while maintaining high detection accuracy.

[0074] Ultimately, the method provided in this application first utilizes directional illumination under a first illumination mode, enabling the system to efficiently perform preliminary registration mark positioning and deviation calculation. When a deviation exceeds a threshold, instead of immediately triggering an alarm or making adjustments, the system further assesses the confidence level of the deviation value, allowing the system to self-check the reliability of the preliminary detection results. When the confidence level is insufficient to support the current deviation result, the system intelligently triggers diffuse illumination under a second illumination mode and acquires a second reflection image. By analyzing the surface texture features in the second reflection image, the system can acquire more targeted information, thereby determining whether suspicious abnormal states are caused by interference from surface deposits. Compared to traditional systems that cannot distinguish true deviations, this application can not only detect color misalignment but also detect the potential causes of the deviation, thus avoiding numerous false alarms caused by surface deposit interference in traditional methods. This allows operators to perform accurate problem troubleshooting and handling, reducing unplanned downtime and improving production efficiency.

[0075] Furthermore, in some embodiments, the method for performing step S300, which assesses the confidence level of the color misalignment value based on the first reflection image, will include the following steps:

[0076] S310. Extract various visual evaluation features related to the registration marks from the first reflection image;

[0077] S320. Normalize the score for each visual evaluation feature;

[0078] S330. The normalized scores are combined using a weighted geometric mean algorithm to obtain a confidence level within a set range.

[0079] This scheme aims to extract multi-dimensional information related to registration marks from the first reflection image to comprehensively characterize the visual quality and reliability of the registration marks. These features can be quantitative indicators reflecting mark sharpness, integrity, and contrast with the background. By extracting multiple features, the limitations of single features can be avoided, improving the accuracy and robustness of the registration mark status assessment. Subsequently, visual evaluation features of different types and scales are normalized to a comparable scale for subsequent comprehensive processing. Normalization typically maps the original feature values ​​to a preset numerical range, allowing the contribution of different features to be measured under a unified standard. This can be achieved through various methods such as linear mapping, the Sigmoid function, and Z-score normalization, ensuring the fairness and comparability of the scores. Finally, the scores of the multiple normalized visual evaluation features are fused into a single confidence score representing the reliability of the color deviation value. In this fusion assessment process, the weighted geometric mean algorithm effectively handles the non-linear relationship between different features, and different features are assigned corresponding weights, which can be adjusted according to their importance to the confidence score assessment, thereby generating a more robust and reliable comprehensive score. Ultimately, the obtained confidence values ​​are limited to a preset range, such as 0 to 1, 1 to 100, etc., where higher values ​​indicate that the color deviation value has higher confidence.

[0080] As a specific implementation method, in some embodiments provided in this application, the extraction of various visual evaluation features related to the registration mark from the first reflection image will specifically include: edge sharpness features, template matching confidence features, region contrast features, and positional residual features.

[0081] The edge sharpness feature is a representation of the clarity of the registration mark edges. This feature aims to assess image quality and mark visibility, and its value can be affected by image blur, noise, or surface interference. For example, it can be obtained by calculating the magnitude or consistency of the gradient direction of the pixel grayscale gradients within the registration mark region, such as using Sobel, Prewitt, or Canny operators to detect edges and calculating the average gradient magnitude of edge pixels. Another approach is to analyze the steepness of the grayscale profile curve of the registration mark edges; the steeper the curve, the sharper the edge.

[0082] The template matching confidence feature is a score calculated by combining multiple related peak morphology information from the related peaks generated after matching a preset registration mark template with the first reflection image. This feature is used to evaluate the reliability of the template matching result to distinguish between true and false matches. The multiple related peak morphology information involved includes at least two of the following: peak sharpness, signal-to-noise ratio (SNR) of the related peak, and the intensity ratio of the main peak to the secondary peak. In this embodiment, three of these are preferred. Peak sharpness can be quantified by calculating the ratio of the grayscale value of the 3x3 pixel region surrounding the main peak to the main peak value; higher sharpness indicates more accurate matching. The SNR of the related peak is the ratio of the main peak value to the standard deviation of the grayscale value of the non-peak region on the related plane; a higher SNR means the matching result is less affected by noise. The intensity ratio of the main peak to the secondary peak is directly calculated as the ratio between the main peak value and the second highest peak value; a larger ratio indicates a unique and definite matching location, reducing the possibility of multiple matches. After obtaining the three related peak morphology information items, these three items need to be combined and analyzed to obtain the template matching confidence feature. In some embodiments, a comprehensive score can be obtained, for example, through a preset function or lookup table; in this embodiment, it is preferable to first normalize each item to the [0, 1] interval to obtain a normalized value, where the closer the value is to 1, the more reliable the indicator is. Then, the three items are calculated using a preset weighted geometric mean formula, and the template matching confidence value is used to participate in subsequent calculations. The specific weights of the three items can be obtained by the operator based on experience or by pre-calibration.

[0083] Region contrast features are values ​​that represent the degree of grayscale difference between the registration mark region and the background region. This feature aims to evaluate the distinguishability of the registration mark and the background, as insufficient contrast directly affects positioning accuracy. This feature can be obtained by calculating the absolute value of the difference between the average grayscale value of the registration mark region and the average grayscale value of the adjacent background region. Alternatively, it can be quantified by calculating the overlap or Kullback-Leibler divergence of the grayscale histograms of the registration mark region and the background region; lower overlap or higher divergence indicates better contrast.

[0084] The position residual feature is a representation of the deviation between the current calculated position of the registration mark and the expected position predicted based on historical position data. This feature is used to detect anomalous jumps in the registration mark position, thereby distinguishing between real motion and transient interference. This feature can be achieved by establishing a Kalman filter or linear regression model, fitting a linear trend line using the least squares method based on the registration mark position data of the previous N frames (e.g., the previous 10 frames), predicting the expected position of the current frame, and then calculating the Euclidean distance between the detected position and the predicted position. Alternatively, the difference between the current detected position and the detected position of the previous frame can be calculated and compared with the historical average position change trend; if it exceeds the statistical range, the residual is considered large.

[0085] Finally, when calculating the confidence score based on the aforementioned multiple feature parameters, for example, in this embodiment, the edge sharpness feature is assigned a weight of 0.3, the template matching confidence feature is assigned a weight of 0.4, the region contrast feature is assigned a weight of 0.2, and the location residual feature is assigned a weight of 0.1. Then, these scores are calculated using a geometric mean according to their respective weights, i.e.:

[0086] Confidence score = (Edge sharpness score^0.3 * Template matching confidence feature score^0.4 * Region contrast feature score^0.2 * Location residual feature score^0.1)^(1 / (0.3+0.4+0.2+0.1))

[0087] The final confidence score will be a value between 0 and 100, used to determine the reliability of the color registration deviation. In other feasible methods, the above-mentioned multiple feature parameters can also be weighted differently based on the experience of technicians or relevant calibration models, which will not be elaborated here.

[0088] Through the above technical solution, this application provides a structured and multi-dimensional confidence assessment method, which significantly improves the accuracy and reliability of color misalignment value assessment. By extracting multiple visual assessment features, it can capture the visual information of the registration mark from a more comprehensive perspective, effectively avoiding the one-sidedness that may be caused by single feature assessment. Among these visual assessment features, the edge sharpness feature effectively quantifies the clarity of the registration mark edge; if the edge is blurry, it may indicate poor image quality or interference. The template matching confidence feature analyzes the relevant peak shape information of the template matching result, such as sharpness, signal-to-noise ratio, and the intensity ratio of the main peak to the secondary peak, to judge the reliability of the matching and avoid the impact of erroneous positioning caused by false matching. The region contrast feature assesses the grayscale difference between the registration mark and the background to ensure that the mark has sufficient recognizability in the image. The position residual feature uses historical data to predict the expected position of the mark and compares it with the current detection position to identify abnormal position jumps caused by instantaneous interference rather than real mechanical movement. At the same time, these preferred features do not work in isolation, but complement each other and work synergistically. For example, when edge sharpness is low, template matching confidence is low, regional contrast is poor, and positional residuals are large, the system will comprehensively determine that the confidence level of the current color registration deviation value is low. Ultimately, this comprehensive evaluation mechanism enables the system to fully consider the reliability of the color registration deviation value from multiple dimensions such as image quality, matching reliability, marker recognizability, and positional stability.

[0089] Furthermore, in some embodiments, the method for determining whether a suspected abnormal state is caused by interference from deposits on the surface of the printed material based on the second reflection image in step S520 includes the following steps:

[0090] S521. Determine the analysis area in the second reflection image based on the position of the registration mark in the first reflection image;

[0091] S522. Calculate at least two discriminant features of the analysis region; the discriminant features are selected from the pixel window unit texture contrast of the analysis region, the shape irregularity of the gray-level binary region in the analysis region, and the structural similarity between the analysis region and the corresponding region in the preset historical cleaning reference image of the printing material;

[0092] S524. Based on at least two discriminative features, determine whether the suspicious abnormal state is caused by interference from surface deposits.

[0093] The above scheme first determines the analysis area in the second reflection image by identifying the registration mark position in the first reflection image, aiming to focus subsequent analysis on specific areas on the printed material where surface deposits may exist. In practice, the center coordinates or bounding box of the registration mark can be accurately identified in the first reflection image. Since the first and second reflection images are acquired continuously within the same detection cycle, and the camera position is relatively fixed, the registration mark position in the first reflection image can be accurately mapped to the corresponding area in the second reflection image using pre-calibrated camera parameters and image registration algorithms.

[0094] Next, this scheme calculates at least one discriminative feature of the analysis region. These discriminative features are selected from the pixel window unit texture contrast of the analysis region, the shape irregularity of the gray-level binary region within the analysis region, and the structural similarity between the analysis region and a preset historical cleaning reference image of the printed material. This step aims to extract quantitative indicators from the analysis region of the second reflection image that can effectively distinguish surface deposits from normal printed material.

[0095] The choice of pixel window unit texture contrast reflects the drastic degree of local image grayscale changes within the analysis area. This is because the second illumination mode provides uniform diffuse illumination, and the texture contrast of normal printed materials is usually relatively stable. However, surface deposits (such as dust and oil) alter local reflectivity, causing a significant difference in texture contrast between their area and the surrounding normal area. In practice, a sliding window approach can be used, moving a small pixel window (e.g., 5x5 or 7x7 pixels) within the analysis area with a preset step size. The grayscale standard deviation or local entropy value within each window is calculated as the texture contrast. These local contrast values ​​can be further analyzed to determine their average, maximum, or distribution characteristics as a basis for judgment.

[0096] The choice of shape irregularity for the grayscale binary regions leverages the fact that real registration marks typically have regular geometric shapes, while surface attachments are random and irregular in shape. By performing grayscale thresholding on the analysis area and binarizing the image, potential attachment regions can be obtained. Then, the shape factors of these binary regions are calculated, such as circularity (4π * area / perimeter^2) or rectangularity (area / area of ​​the smallest bounding rectangle). The closer the circularity is to 1, the closer the shape is to a circle; the closer the rectangularity is to 1, the closer the shape is to a rectangle. Attachments typically exhibit low circularity or rectangularity, i.e., irregular shapes.

[0097] The selection of structural similarity between the analysis region and a preset historical clean reference image of the printed material is achieved by directly quantifying the structural changes of the current analysis region through comparison with the reference image under a known clean state. The preset historical clean reference image of the printed material is an image acquired when the printed material is in a clean and undisturbed state, representing the intrinsic texture of the material. When surface deposits are present in the analysis region, its structure changes, leading to a decrease in similarity with the reference image. In implementation, the Structural Similarity Index (SSIM) algorithm or the Normalized Cross-Correlation (NCC) algorithm can be used to calculate the similarity score between the current analysis region and the corresponding reference image region. A lower similarity score indicates the presence of interference.

[0098] Finally, based on at least two discriminative features, a comprehensive judgment can be made as to whether a suspicious abnormal state is caused by surface attachment interference. This allows for the integrated use of information from multiple discriminative features, improving the robustness and accuracy of the judgment and avoiding misjudgments that may result from a single feature. During the judgment process, thresholds or threshold ranges corresponding to multiple discriminative features can be set. For example, if the texture contrast of the pixel window unit in the analysis area exceeds a preset pixel window unit texture contrast threshold, and the shape irregularity of the grayscale binary region is lower than a preset lower limit for shape irregularity, then it is determined to be surface attachment interference. Alternatively, a machine learning classifier, such as a support vector machine (SVM) or decision tree, can be used, taking at least two extracted discriminative features as input and outputting the judgment result through a pre-trained model.

[0099] The solution of this application introduces image analysis under a second illumination mode. Relying on the uniform diffuse illumination of the second illumination mode, it can truly reflect the texture and attachment characteristics of the printed material surface. Then, in the case of suspected abnormality, it uses the multi-dimensional discrimination features in the second reflection image acquired under the second illumination mode to make a comprehensive judgment. The system can accurately distinguish between real color registration deviation and false deviation caused by surface attachments such as dust and oil stains.

[0100] Furthermore, considering that the combinations of discriminative features obtained from the analysis may vary significantly depending on the type of deposits present in the analysis area, specifically, when the pixel window unit texture contrast is high and the shape irregularity of the gray-scale binary region is low, it is more likely to indicate hard particles with clear boundaries (such as metal fragments) rather than soft stains; while when the pixel window unit texture contrast is low and the structural similarity is low, it may indicate large-area, uniform deposits (such as a thin oil film), based on the fact that it changes the overall reflectivity but does not increase the local texture complexity; however, if the pixel window unit texture contrast is at a high level and the shape irregularity is at a high level, while the structural similarity is at a low level, it strongly and typically points to random surface deposits, such as dust clumps or splatter.

[0101] Based on the above, in order to more accurately use the discriminative features to determine the attachments, this embodiment will, based on the combination relationship between these feature values ​​and the correspondence in typical scenarios, in step S524, not only check whether each feature exceeds its independently set threshold range, but also combine and calculate the specific parameters of the above three discriminative features and analyze the situation.

[0102] Specifically, firstly, a comprehensive score is calculated based on three discriminant features. This comprehensive score S reflects the overall "degree of abnormality," and the calculation formula is as follows:

[0103] Overall score S = w1 * N(F1) + w2 * [1-N(F2)] + w3 * [1-N(F3)]

[0104] In the formula, N(.) represents the normalization of the feature values; F1 is the texture contrast of the pixel window unit; F2 is the shape irregularity of the gray-scale binary region; F3 is the structural similarity between the analysis region and the preset historical cleaning reference image of the printed material; w1, w2, and w3 are the weights of each discriminative feature, and the sum of the weights is 1, where w3 is the highest, for example 0.5, w2 is the second highest, for example 0.3, and w1 is relatively low, for example 0.2.

[0105] After calculating the overall score S, if S is greater than the set overall score threshold, it is preliminarily determined that there is interference. This embodiment will further distinguish the type of interference and will specifically combine the following branches for judgment:

[0106] If F2 is lower than the preset lower limit threshold for irregular shape of gray-scale binary region, it indicates that the shape of the current region is very irregular, fragmented or extremely complex. If the comprehensive score is too high, it will be finally judged as a typical random attachment.

[0107] If F2 is higher than the upper limit threshold of irregular shape of gray-scale binary region preset, and F3 is lower than the preset structural similarity threshold, it indicates that the shape of the current region is close to circle or relatively regular, which is inconsistent with the typical characteristics of random attachments. Combined with the extremely low F3, it is judged as large-area uniform pollution.

[0108] This embodiment utilizes a multi-discriminative feature fusion approach to construct a more robust and intelligent discrimination model by mining the inherent correlation patterns among three complementary features: texture, shape, and historical comparison. This not only improves the recall rate for detecting deposits on printed materials, but more importantly, by analyzing feature combination patterns, it significantly reduces the false positive rate. This allows the system to more accurately distinguish between "surface deposit interference" and "true color registration deviation," thus providing a crucial guarantee for achieving ultra-high reliability in color registration deviation detection.

[0109] Furthermore, for thin films, composite laminates, and similar materials, the microscopic deformation or uneven distribution of internal stress under tension and temperature can lead to non-rigid deformation in the registration mark area even without any attached material. This manifests as an anomaly in a single image that is difficult to distinguish from the actual color misalignment. In this case, even the second lighting mode is insufficient for effective identification. Based on this problem that is prone to occur in this specific type of material, this application further proposes in some embodiments that, after step S400, when the confidence level is lower than a preset confidence level threshold, the condition is determined to be a suspicious anomaly, the following steps are also included:

[0110] S610. In response to a suspected abnormal state, acquire at least two third reflection images of the printed material under the illumination of a third illumination mode; wherein the third illumination mode is a linearly polarized light source providing polarized light illumination to the surface of the printed material in the same direction as the first illumination mode, the third reflection images are captured by a camera combined with a polarization state analyzer, and the two third reflection images have different polarization state configurations.

[0111] S620. Based on the two third reflection images, calculate the polarization feature information related to the optical anisotropy of the material; the polarization feature information is a polarization degree image or a polarization difference image.

[0112] S630. Based on the polarization characteristic information, determine whether the suspected abnormal state is caused by internal stress or deformation of the printing material;

[0113] S640. If so, output a prompt signal indicating that the color deviation value is caused by internal stress or deformation of the printing material.

[0114] The acquisition of at least two third reflection images of the printed material under a third illumination mode aims to introduce a new imaging mode for detecting the internal properties of the printed material. This can be achieved by integrating the light source and corresponding control unit of the third illumination mode into a visual prediction system for color misregistration in a multi-color printing press, and integrating a polarization state analyzer and camera into the imaging device, so that these images can be triggered and acquired in a timely manner when the system identifies a suspicious abnormal state. Alternatively, a separate polarization imaging module can be used, which works in conjunction with the main visual prediction system to acquire polarization images of specific areas of the printed material after receiving a suspicious abnormal state signal. The third illumination mode is a linearly polarized light source that provides polarized light illumination to the surface of the printed material in the same direction as the first illumination mode. Its function is to ensure that the illumination light has a specific polarization direction, and that this direction is consistent with the direction of the first illumination mode used to highlight the geometric edges of the registration marks. This helps in subsequent analysis of the material's response to polarized light and maintains the correlation with the geometric edge imaging mode. This can be achieved by using an LED array light source and setting a linear polarizer in its light output path. The transmission axis of the polarizer needs to be pre-calibrated to be consistent with the direction of the first illumination mode. Another approach is to use a laser diode as the light source, which itself outputs linearly polarized light, and adjusts its polarization direction via a rotation mechanism. The third reflection image is captured by a camera incorporating a polarization state analyzer, and the two third reflection images have different polarization state configurations. The purpose is to capture reflected light with different polarization states through the analyzer to reveal the optical anisotropy of the material. Different polarization state configurations are crucial for calculating polarization characteristic information. This can be achieved using an industrial camera with a rotatable polarizer (i.e., a polarization state analyzer), capturing two images by rotating the polarizer to at least two different angles.

[0115] Subsequently, based on the two third reflection images, polarization feature information related to the optical anisotropy of the material is calculated. The polarization feature information is a polarization degree image or a polarization difference image, which is used to extract information on the internal structural properties of the material from images of different polarization states, such as stress and deformation.

[0116] For polarization images, the polarization degree can be calculated pixel-by-pixel using the formula: P = (I_parallel - I_perpendicular) / (I_parallel + I_perpendicular) based on the intensities I_parallel and I_perpendicular of two images with different polarization states (e.g., orthogonal polarization states). The calculated polarization degree P is a dimensionless ratio with a range of [-1, 1]. It represents the degree of order in the vibration direction of light waves in a specific detection direction. P ≈ 0 indicates that the light reflected or transmitted from that point in the material is unpolarized (or circularly polarized), meaning that the optical properties of the material at that point are isotropic, with a uniform internal structure and no significant stress-induced directional alignment or birefringence. |P| > 0 indicates that the light has a linear polarization component; the larger the absolute value, the stronger the anisotropic response of the material to incident polarized light at that point. This usually directly corresponds to the optical anisotropy (i.e., birefringence effect) caused by stress, deformation, or molecular orientation within the material.

[0117] For polarization difference images, two images with different polarization states can be directly subtracted pixel by pixel, i.e., D = I_parallel - I_perpendicular.

[0118] The polarization difference D is a difference image with intensity units. It directly reflects the absolute difference in a material's ability to reflect or transmit two orthogonally polarized lights. When D ≈ 0, it indicates that the material's response to the two polarized lights is almost the same, suggesting isotropy; a large |D| value highlights regions of strong anisotropy within the material. These regions may not be obvious in a single intensity image, but are significantly enhanced by subtraction.

[0119] Subsequently, when determining whether the suspected abnormal state is caused by internal stress or deformation of the printing material based on the polarization feature information, image processing algorithms can be used to analyze whether specific patterns exist in the polarization feature image, such as regular interference fringes, birefringent regions, or abnormal polarization degree distributions. These patterns are usually associated with internal stress or deformation of the material. Alternatively, a machine learning model can be trained, taking the polarization feature image as input and outputting a judgment result. This model establishes classification rules by learning from a large number of polarization feature images with and without internal stress / deformation samples.

[0120] The above-described solution introduces polarized light imaging technology, adding a diagnostic step in cases of suspected anomalies, thus solving the problem of being unable to identify anomalies caused by internal stress or deformation. In this way, this solution, combined with basic color misregistration detection and surface deposit interference judgment mechanisms, forms a multi-layered, intelligent diagnostic system. When basic detection detects anomalies with low confidence, it no longer simply alarms or ignores them, but further analyzes them to investigate surface deposit interference and diagnose internal stress or deformation. This significantly improves the accuracy and reliability of color misregistration diagnosis, avoiding false alarms or missed detections caused by a single imaging mode's inability to identify complex anomalies.

[0121] Furthermore, in this embodiment, in the two third reflection images with different polarization state configurations, the transmission axis direction of the polarization state analyzer corresponding to the two third reflection images forms an angle of 0 degrees and 90 degrees with the polarization direction of the linearly polarized light source, respectively.

[0122] Specifically, a polarization state analyzer can be a linear polarizer whose transmission axis direction can be adjusted via a mechanical or electric rotation mechanism. As another possible implementation, the polarization state analyzer can also be a tunable polarizer based on liquid crystal technology, where the alignment of liquid crystal molecules is changed by applying voltage, thereby achieving electrically controlled adjustment of the transmission axis direction. The transmission axis direction refers to the specific direction through which the electric vector vibration direction of the light wave is allowed to pass by the polarization state analyzer; it determines the polarization direction of the light emitted from the polarization state analyzer and is used to control the polarization imaging contrast. A linearly polarized light source is a light source whose emitted light wave electric vector vibrates within a fixed plane. A linearly polarized light source can be implemented by placing a linearly polarizer with a fixed orientation in front of a common light source, or it can itself be a laser or a specific type of LED that emits linearly polarized light. The polarization direction refers to the direction of electric vector vibration of the linearly polarized light emitted by the linearly polarized light source; it defines the initial polarization state of the light incident on the surface of the printed material. Therefore, the fact that the transmission axis direction forms an angle of 0 degrees and 90 degrees with the polarization direction respectively means that the transmission axis direction of the polarization state analyzer is parallel (0 degrees) or perpendicular (90 degrees) to the polarization direction of the linearly polarized light source.

[0123] This application optimizes the polarization image acquisition process by specifically defining the angle between the transmission axis of the polarization state analyzer and the polarization direction of the linearly polarized light source. This ensures efficient capture of the optical anisotropy characteristics of the material, thereby improving the detection accuracy of internal stress or deformation. Specifically, in a 0-degree angle configuration, the polarization state analyzer is parallel to the polarization direction of the linearly polarized light source. In this case, the camera mainly receives the light component that retains its original polarization state after specular reflection from the surface of the printed material. This configuration is sensitive to surface features such as the smoothness of the material surface and the thickness of the film layer, and can enhance the visibility of surface reflection features. In a 90-degree angle configuration, the polarization state analyzer is orthogonal to the polarization direction of the linearly polarized light source. In this case, the polarization state analyzer effectively suppresses specular reflection light from the surface of the printed material, and mainly receives the light components whose polarization direction has changed after multiple scattering, refraction, or birefringence due to internal stress. This orthogonal polarization configuration is highly sensitive to internal features such as internal stress, crystal structure, and subsurface defects of the material, and can effectively suppress surface reflection and highlight birefringence caused by internal stress. By combining this 0-degree and 90-degree angle, the two acquired third-reflection images physically carry almost orthogonal and complementary optical information. This complementarity allows for the significant suppression of common background information unrelated to polarization, such as uniform ink color or dust shadows, when subsequently calculating polarization feature information (such as polarization degree images or polarization difference images). Therefore, high-contrast polarization feature image data can be generated, facilitating the clear identification of interference fringes or birefringence features caused by material anisotropy.

[0124] Furthermore, in this embodiment, when determining whether a suspected abnormal state is caused by internal stress or deformation of the printing material based on polarization characteristic information in step S630, the following steps are included:

[0125] S631. Analyze whether the polarization characteristic information presents regular interference fringes or significant birefringence characteristics in the region where the registration mark is located;

[0126] S632. If so, the suspected abnormal state is determined to be caused by internal stress or deformation of the material.

[0127] S633. If not, then the suspected abnormal state is determined to be not caused by internal stress or deformation of the material.

[0128] In the above scheme, polarization degree images or polarization difference images can reflect the optical anisotropy of materials. Therefore, when stress or deformation exists within the material, its molecular structure changes, causing the material to exhibit a birefringence effect. That is, light with different polarization directions travels at different speeds within the material, thus forming specific interference fringes or birefringence features in the polarization image. Specifically, frequency domain analysis methods such as Fourier transform or wavelet transform can be used to detect whether periodic interference fringes exist in the image. Alternatively, a preset pattern recognition algorithm can be used to match the collected polarization feature information with templates of known stress or deformation patterns to determine whether regular interference fringes or significant birefringence features exist. If so, the suspected abnormal state is determined to be caused by internal stress or deformation of the material.

[0129] When polarization feature analysis clearly shows regular interference fringes or significant birefringence, it directly indicates the presence of stress or deformation within the printing material. These internal defects alter the material's optical properties, affecting the visual appearance of registration marks and leading to anomalies in the calculation of color misalignment. Therefore, the suspected anomaly is attributed to internal material stress or deformation. If not, the suspected anomaly is determined not to be caused by internal material stress or deformation, achieving a process of elimination. If polarization feature analysis does not show regular interference fringes or significant birefringence, internal material stress or deformation can be ruled out as the cause of the current suspected anomaly. This helps narrow down the diagnostic scope to other potential interfering factors, such as surface deposits, thereby improving the efficiency and accuracy of fault diagnosis.

[0130] Furthermore, in some embodiments, when the confidence level is lower than a preset confidence level threshold, the second reflection image is acquired within a set duration threshold or the second reflection image and the third reflection image are acquired continuously; the duration threshold is calculated based on the conveying speed of the printing material and the length of the area where the registration mark is located in the conveying direction of the printing material.

[0131] Specifically, the first reflection image, the second reflection image, and the optional third reflection image are the basic data for subsequent analysis and judgment. These three images must be acquired sequentially in a preset order within a very short timeframe as the printed material passes the detection location, ensuring that images under different lighting conditions correspond to approximately the same physical location on the material during its movement.

[0132] To ensure continuous image acquisition under the aforementioned multiple modes and for the same physical location on the printing material, this application solution preferably utilizes a lighting device and shooting device capable of achieving the three modes.

[0133] Specifically, the lighting device employs an integrated light source module with electrically switchable polarization states. The core of this module includes a light-emitting unit (such as an LED array) and an optically integrated liquid crystal polarization control unit. By applying different driving voltages to the liquid crystal unit, it can switch between two optical states in microseconds.

[0134] In the first state, the unit exhibits optical isotropy, exhibiting no polarization selection on the light emitted by the light-emitting unit, thus outputting unpolarized basic illumination light. This state is used as the initial light source for the first and second illumination modes. In the second state, the unit functions as a polarizer, converting the light emitted by the light-emitting unit into linearly polarized light with a specific vibration direction. This state is dedicated to the third illumination mode. A polymer dispersed liquid crystal (PDLC) dimming panel is further coupled in front of the light source module. This PDLC panel can switch between a transparent state and a hazy (scattering) state in milliseconds under the control of the driving voltage. In the transparent state, the panel maintains high transmittance for incident light (whether basic or polarized), suitable for the first and third illumination modes requiring high directionality. In the hazy state, the dimming panel itself becomes a uniform secondary light-emitting surface, scattering incident light to form uniform diffuse illumination, dedicated to the second illumination mode.

[0135] The imaging device employs a high frame rate industrial camera, with a liquid crystal polarization rotator integrated in front of its lens as a polarization analyzer. This liquid crystal polarization rotator can also control its optical state via a driving voltage: in the inactive state, it exhibits high transmittance across all incident light polarization directions, acting as a transparent window and not affecting the acquisition of the first and second reflection images; in the active state, it functions as an electrically tunable analyzer, allowing only polarized light with a set vibration direction to pass through, specifically for analyzing the polarization state of the reflected light under the third illumination mode.

[0136] The system's operating sequence is as follows: When the system determines that an image needs to be acquired in a certain mode, the main controller synchronously sends drive signals to all the aforementioned components. For example, when the first illumination mode is triggered, the light source module is controlled to output unpolarized light, the PDLC panel is switched to a transparent state, and the camera polarization rotator is placed in an inactive state, followed by camera exposure. If switching to the second illumination mode is required, simply controlling the PDLC panel to switch to a fog state is sufficient for image acquisition. If switching to the third illumination mode is required, the light source module can be synchronously controlled to output linearly polarized light, the PDLC panel remains transparent, the camera polarization rotator is activated and tuned to a predetermined polarization detection angle, followed by exposure. The state switching of all electro-optical components can be completed and stabilized in milliseconds or even less, thus ensuring high-speed, lossless continuous acquisition of images of the same physical location under different illumination modes within a single detection cycle, providing a crucial data foundation for subsequent accurate comparison and intelligent diagnosis.

[0137] Furthermore, the time interval between two adjacent images in the acquisition sequence is no greater than a set duration threshold. This aims to strictly limit the time difference between consecutive image acquisitions, ensuring that the physical position offset of adjacent images on high-speed conveyed printing material remains within an acceptable range. This duration threshold is a dynamic parameter, calculated based on the conveying speed of the printing material and the length of the registration mark area along the material's conveying direction. In this embodiment, a baseline time is preferably obtained by dividing the length of the registration mark by the conveying speed of the printing material, and then the duration threshold is set as a fraction of this baseline time or a fixed value slightly smaller than it. This dynamic calculation ensures that the synchronization of image acquisition is effectively guaranteed under different production speeds and mark sizes.

[0138] The duration threshold setting ensures that all images capture nearly the same area on the printed material, thus providing highly aligned and reliable image data for subsequent color registration deviation calculations, confidence assessments, and diagnosis of suspicious abnormalities.

[0139] Secondly, this application proposes a visual prediction device for color registration deviation in multicolor printing presses.

[0140] Reference Figure 2 This is a schematic diagram of a visual prediction device for color registration deviation in a multicolor printing press proposed in this application. The device includes an image acquisition module, a color registration deviation calculation module, and a color registration deviation result judgment module.

[0141] The image acquisition module is configured to acquire a first reflected image of the printed material under a first illumination mode; wherein, under the first illumination mode, the light source provides directional illumination to the surface of the printed material to highlight the geometric edges of the registration marks on the printed material. In response to a suspected abnormal state, the image acquisition module also acquires a second reflected image of the printed material under a second illumination mode; under the second illumination mode, the light source provides uniform diffuse illumination to the surface of the printed material to highlight the surface texture of the printed material.

[0142] The color misalignment calculation module locates the registration mark and calculates the color misalignment value based on the first reflection image;

[0143] If the color registration deviation result judgment module exceeds the preset deviation threshold, it evaluates the confidence level of the color registration deviation value based on the first reflection image; if the confidence level is lower than the preset confidence threshold, it is determined to be a suspicious abnormal state; based on the second reflection image, it determines whether the suspicious abnormal state is caused by interference from the surface of the printing material; if so, it outputs a prompt signal indicating that the color registration deviation value is caused by surface interference of the printing material.

[0144] Furthermore, in some embodiments, the image acquisition module is also configured to, after determining a suspicious abnormal state when the confidence level is lower than a preset confidence level threshold, perform the following steps:

[0145] At least two third reflection images of the printed material are acquired under the illumination of a third illumination mode; wherein the third illumination mode is a linearly polarized light source that provides polarized light illumination to the surface of the printed material in the same direction as the first illumination mode, the third reflection images are captured by a camera combined with a polarization state analyzer, and the two third reflection images have different polarization state configurations.

[0146] The color registration deviation result judgment module is also configured to calculate polarization feature information related to the optical anisotropy of the material based on the two third reflection images; the polarization feature information is a polarization degree image or a polarization difference image; based on the polarization feature information, it is determined whether the suspected abnormal state is caused by internal stress or deformation of the printing material; if so, a prompt signal is output that the color registration deviation value is caused by internal stress or deformation of the printing material.

[0147] Thirdly, this application proposes a visual prediction system for color registration deviation in multi-color printing presses, which includes:

[0148] An illumination device includes a light source and a PDLC transmissive plate disposed in the light path between the light source and the printed material. The PDLC transmissive plate can switch between a transparent state and a frosted state. The illumination device provides the printed material with a light mode including at least a first illumination mode and a second illumination mode through the PDLC.

[0149] An imaging device for capturing a corresponding reflected image of the printed material illuminated by the lighting device;

[0150] The visual prediction device for color registration deviation in multicolor printing presses as described above.

[0151] Furthermore, in some embodiments, the core of the light source includes a light-emitting unit (such as an LED array) and an optically integrated liquid crystal polarization control unit. By applying different driving voltages to the liquid crystal polarization control unit, it can switch between two optical states on a microsecond-level basis: In the first state, the liquid crystal polarization control unit is optically isotropic and does not produce polarization selection on the light emitted by the light-emitting unit, thereby outputting unpolarized basic illumination light. This state is used as the initial light source for the first illumination mode and the second illumination mode; In the second state, the liquid crystal polarization control unit operates as a polarizer, converting the light emitted by the light-emitting unit into linearly polarized light with a specific vibration direction. This state is dedicated to the third illumination mode.

[0152] The imaging device employs a high frame rate industrial camera, with a liquid crystal polarization rotator integrated in front of its lens as a polarization analyzer. This liquid crystal polarization rotator can also control its optical state via a driving voltage: in the inactive state, it exhibits high transmittance across all incident light polarization directions, acting as a transparent window and not affecting the acquisition of the first and second reflection images; in the active state, it functions as an electrically tunable analyzer, allowing only polarized light with a set vibration direction to pass through, specifically for analyzing the polarization state of the reflected light under the third illumination mode.

[0153] The system's operating sequence is as follows: When the system determines that an image needs to be acquired in a certain mode, the main controller synchronously sends drive signals to all the aforementioned components. For example, when the first illumination mode is triggered, the system controls the light source to output unpolarized light, switches the PDLC panel to a transparent state, and deactivates the liquid crystal polarization rotator in the camera, subsequently triggering camera exposure. To switch to the second illumination mode, simply switch the PDLC panel to a fog state for image acquisition. To switch to the third illumination mode, the system synchronously controls the light source to output linearly polarized light, keeps the PDLC panel transparent, activates the polarization rotator, and tunes it to a predetermined polarization detection angle, followed by exposure. The state switching of all electro-optical components can be completed and stabilized within milliseconds or even less, thus ensuring high-speed, non-destructive continuous acquisition of images of the same physical location under different illumination modes within a single detection cycle, providing a crucial data foundation for subsequent accurate comparison and intelligent diagnosis.

[0154] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0155] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0156] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0157] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0158] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0159] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0160] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A visual prediction method for color registration deviation in a multi-color printing press, characterized in that, It includes the following steps: Acquire a first reflected image of the printed material under a first illumination mode; wherein, under the first illumination mode, the light source provides directional illumination to the surface of the printed material to highlight the geometric edges of the registration marks on the printed material; Based on the first reflection image, locate the registration mark and calculate the color deviation value; If the color mismatch value exceeds a preset mismatch threshold, the confidence level of the color mismatch value is evaluated based on the first reflection image; When the confidence level is lower than the preset confidence level threshold, it is determined to be a suspicious abnormal state; In response to the suspected abnormal state, a second reflection image of the printed material under the illumination of the second lighting mode is acquired; wherein, under the second lighting mode, the light source provides uniform diffuse illumination to the surface of the printed material to highlight the surface texture of the printed material. Based on the second reflection image, determine whether the suspected abnormal state is caused by interference from substances adhering to the surface of the printed material; If so, output a prompt signal indicating that the color mismatch value is caused by surface interference of the printing material.

2. The visual prediction method for color registration deviation in multi-color printing presses as described in claim 1, characterized in that, The step of evaluating the confidence level of the color mismatch value based on the first reflective image includes the following steps: Extract various visual evaluation features related to the registration mark from the first reflection image; A normalized score is applied to each of the aforementioned visual evaluation features; The normalized scores are combined using a weighted geometric mean algorithm to obtain a confidence level within a set range.

3. The visual prediction method for color registration deviation in multi-color printing presses as described in claim 2, characterized in that, The multiple visual evaluation features include at least two of edge sharpness features, template matching confidence features, region contrast features, and location residual features; Wherein, the edge sharpness feature is a characterization value reflecting the clarity of the registration mark edge; The template matching confidence feature is a score value calculated by combining multiple related peak morphology information in the related peaks generated after matching the preset registration mark template with the first reflection image; the multiple related peak morphology information includes at least two of the following: the sharpness of the related peak, the signal-to-noise ratio of the related peak, and the intensity ratio of the main peak to the secondary peak; The region contrast feature is a characterization value that reflects the degree of grayscale difference between the region where the registration mark is located and the background region; The position residual feature is a characterization value reflecting the deviation between the position of the registration mark obtained in this calculation and the expected position predicted based on historical position data.

4. The visual prediction method for color registration deviation in a multi-color printing press as described in claim 1, characterized in that, The step of determining whether the suspected abnormal state is caused by interference from substances adhering to the surface of the printed material based on the second reflection image includes the following steps: The analysis region in the second reflection image is determined based on the position of the registration mark in the first reflection image; Calculate at least two discriminant features of the analysis region; the discriminant features are selected from the pixel window unit texture contrast of the analysis region, the shape irregularity of the gray-level binary region in the analysis region, and the structural similarity between the analysis region and the corresponding region in the preset historical cleaning reference image of printing materials; Based on at least two of the discriminative features, it is determined whether the suspected abnormal state is caused by interference from surface deposits.

5. The visual prediction method for color registration deviation in a multi-color printing press as described in claim 1, characterized in that, After determining a suspicious abnormal state when the confidence level is lower than a preset confidence level threshold, the process further includes the following steps: At least two third reflection images of the printed material are acquired under the illumination of a third illumination mode; wherein the third illumination mode is a linearly polarized light source that provides polarized light illumination to the surface of the printed material in the same direction as the first illumination mode, the third reflection images are captured by a camera combined with a polarization state analyzer, and the two third reflection images have different polarization state configurations. Based on the two third reflection images, polarization feature information related to the optical anisotropy of the material is calculated; the polarization feature information is a polarization degree image or a polarization difference image. Based on the polarization characteristic information, determine whether the suspected abnormal state is caused by internal stress or deformation of the printing material; If so, output a warning signal indicating that the color mismatch value is caused by internal stress or deformation of the printing material.

6. The visual prediction method for color registration deviation in a multi-color printing press as described in claim 5, characterized in that, The two third reflection images have different polarization state configurations. The transmission axis of the polarization analyzer corresponding to the two third reflection images forms an angle of 0 degrees and 90 degrees with the polarization direction of the linearly polarized light source, respectively.

7. The visual prediction method for color registration deviation in a multi-color printing press as described in claim 5, characterized in that, The step of determining whether the suspected abnormal state is caused by internal stress or deformation of the printing material based on the polarization feature information includes the following steps: Analyze whether the polarization characteristic information exhibits regular interference fringes or significant birefringence characteristics within the region where the registration mark is located; If so, the suspected abnormal state is determined to be caused by internal stress or deformation of the material; If not, then the suspected abnormal state is determined not to be caused by internal stress or deformation of the material.

8. The visual prediction method for color registration deviation in a multi-color printing press as described in claim 1 or 5, characterized in that, When the confidence level is lower than a preset confidence level threshold, the second reflection image is acquired within a set time threshold, or the second reflection image and the third reflection image are acquired continuously; the time threshold is calculated based on the conveying speed of the printing material and the length of the area where the registration mark is located in the conveying direction of the printing material.

9. A visual prediction device for color registration deviation in a multi-color printing press, characterized in that, It includes: An image acquisition module is configured to acquire a first reflection image of the printed material under illumination in a first lighting mode, and in response to the suspected abnormal state, acquire a second reflection image of the printed material under illumination in a second lighting mode; wherein, in the first lighting mode, the light source provides directional illumination to the surface of the printed material to highlight the geometric edges of the registration marks on the printed material; in the second lighting mode, the light source provides uniform diffuse illumination to the surface of the printed material to highlight the surface texture of the printed material. A color misalignment calculation module is configured to locate the registration mark and calculate the color misalignment value based on the first reflection image; The color registration deviation result judgment module is configured to: if the color registration deviation value exceeds a preset deviation threshold, evaluate the confidence level of the color registration deviation value based on the first reflection image; if the confidence level is lower than the preset confidence threshold, determine it as a suspicious abnormal state; determine whether the suspicious abnormal state is caused by interference from surface deposits of the printing material based on the second reflection image; if so, output a prompt signal indicating that the color registration deviation value is caused by surface interference of the printing material.

10. A visual prediction system for color registration deviation in a multi-color printing press, characterized in that, It includes: An illumination device includes a light source and a PDLC transmissive plate disposed in the light path between the light source and the printed material. The PDLC transmissive plate can switch between a transparent state and a frosted state. The illumination device provides the printed material with a light mode including at least a first illumination mode and a second illumination mode through the PDLC. An imaging device for capturing a corresponding reflective image of the printed material illuminated by the lighting device; The visual prediction device for color registration deviation in a multi-color printing press as described in claim 9.

Images