A method for aligning the position of a color group printing plate

Abstract

The invention relates to the technical field of image processing, and in particular to a method for aligning the position of a color group printing plate, comprising: acquiring a printed image in real time, judging whether there is an overlapping area, and if there is no overlapping area, matching it with a template image to align the position of the printing plate; if there is an overlapping area, performing the following operations: acquiring the actual edge of the printed image, clustering the area within the actual edge, and calculating the sum of the distances from all pixel points of each cluster cluster to the center points of all other cluster clusters, taking the cluster cluster with the smallest sum of distances as the overlapping area, taking the area outside the overlapping area as the non-overlapping area, calculating the offset vectors of the edges of the overlapping area and the non-overlapping area, and then determining the correction vector of the printed image, and aligning the position of the printing plate based on the correction distance. The method can accurately extract the overlapping area and perform precise position alignment of the printing plate.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for aligning the position of a color group printing plate. Background Art

[0002] In today's printing industry, gravure printing has always been a popular printing method. It is famous for its high-quality printing effect and strong applicability. The gravure printing machine can achieve multi-color printing, which is usually achieved by the method of color group printing. Specifically, it is to superimpose multiple different colors of ink on the paper in a specific order. In this process, each color has a corresponding printing plate, and the precise alignment between these printing plates plays a vital role in ensuring the correct superposition of colors. Achieving accurate alignment of the printing plates is of great significance to the quality of printed products.

[0003] In the management of gravure printing production, the role of ERP (Enterprise Resource Planning) system has become increasingly prominent. Through the interaction between the control system of the gravure printing machine and the ERP system, the automated management and refined control of the entire production process can be achieved. Specifically, the control system of the gravure printing machine transmits data such as printing speed, number of meters printed during the shift, cumulative number of meters printed, strength of the drying boxes of each printing group, and tension to the ERP system for management through the Internet of Things technology, and collects the printed image at the printing plate and transmits it to the ERP system. The ERP system can view the printed image in real time, and can enlarge the image to observe its details, and align the printing plate corresponding to the printed image according to the offset of the printed image, thereby ensuring printing quality.

[0004] In a printed image detection method and detection device disclosed in a Chinese patent document with an existing authorization announcement number of CN106296636B, the detection method first collects the printed image, then uses the automatic threshold segmentation method to roughly extract the edge contour of the foreground area to preliminarily outline the general outline of the foreground part of the image, and then further edits the roughly extracted edge contour according to the grayscale difference threshold, so as to accurately extract the edge contour line of the foreground area. Finally, the edge contour line is located by the area positioning method and the straight line positioning method, so as to complete the detection and analysis of the printed image and then evaluate the printing quality.

[0005] When the above-mentioned technical solution is used to locate the edge contour line of the printed image, since the multi-color printing plate set is printed in sequence by multiple printing plates of different colors, overlapping areas are inevitably produced in the printed image. The existence of these overlapping areas poses a challenge to the control of printing quality, because the edge features of the overlapping areas are often blurred. This ambiguity makes it difficult to accurately extract the real-time printing edge corresponding to the printing plate currently being printed during image processing and analysis, and the edge information cannot be accurately obtained, which leads to deviations in the printing plate position alignment process, ultimately affecting the printing quality. Summary of the invention

[0006] In order to solve the problem that the edge of the overlapping area of ​​the printed image cannot be accurately extracted during the multi-color group printing process, which affects the accuracy of the printing plate alignment and the printing quality, the present invention provides a position alignment method for the color group printing plate, comprising:

[0007] Acquire the printed image of each printing plate of the multi-color printing plate set in real time, determine whether there is an overlapping area of ​​the printed image, and if there is no overlapping area, directly match the printed image with the template image to align the position of the printing plate corresponding to the printed image; if there is an overlapping area, perform the following operations:

[0008] Extract multiple edge pixel point sequences of the printed image, take any two edge pixel point sequences that can form a closed figure as a closed edge, calculate the edge similarity between each closed edge and the template image, and take the closed edge with the greatest edge similarity as the actual edge of the printed image;

[0009] Clustering the area within the actual edge of the printed image to obtain multiple clusters, and calculating the sum of distances from all pixel points of each cluster to the center points of all other clusters, taking the cluster with the smallest sum of distances as the overlapping area, and taking the area of ​​the printed image other than the overlapping area as the non-overlapping area;

[0010] The offset vectors of the edges of the overlapping area and the non-overlapping area are calculated in combination with the edges of the template image, the correction vector of the printed image is determined by combining the offset vectors of the edges of the overlapping area and the non-overlapping area, and the printing plate corresponding to the printed image is aligned based on the correction vector.

[0011] The above technical solution can ensure timely monitoring and adjustment of the printing process by acquiring images in real time. By judging whether there are overlapping areas, the situations are divided into two categories for processing. For situations where there are no overlapping areas, directly matching with the template image is a simple and effective alignment method, because the absence of overlapping areas means that the edge features are relatively clear, and matching with the template can quickly and accurately determine whether the printing plate position is correct. For situations where there are overlapping areas, due to the blurred edges, a more sophisticated processing method is required. Specifically, multiple edge pixel point sequences are extracted to fully capture various edge situations that may exist in the printed image. Based on the closed edges, the areas with complete contours in the image are more accurately defined. These areas often correspond to different parts or objects in the printed image. The actual edge of the printed image is determined by calculating the similarity between the closed edge of the printed image and the ideal template image edge. The part that best matches the edge features of the printed image and the template image can be accurately found, which can be used as the main basis for subsequent analysis and processing. Then, cluster analysis is performed on the inner area of ​​the actual edge of the printed image to obtain a specific cluster that meets the characteristics of the overlapping area, that is, the cluster has more interactions with other areas, which may be due to the area formed by the superposition of different color inks. In this way, the overlapping area and the non-overlapping area are accurately divided, which provides a basis for the subsequent targeted treatment of problems in different areas. Furthermore, the method of comprehensively considering the offset of the overlapping area and the non-overlapping area can more comprehensively and accurately reflect the overall deviation between the printed image and the template image, so as to determine a more reasonable correction distance. Aligning the printing plate position based on this correction distance can effectively correct the inaccurate printing plate position caused by edge deviation, and can make the printing plate position more accurate, thereby reducing printing quality problems such as color deviation and pattern misalignment, and improving the matching degree between the printed product and the template image and the overall printing quality.

[0012] Preferably, the correction vector of the printed image satisfies the following relationship:

[0013]

[0014] In the formula, is the correction vector for the printed image, By printing images respectively , Correction distance for direction , constitute, , are the average color difference between the edge and background colors of the overlapping area and the non-overlapping area, , They are the edge similarities between the overlapping area, the non-overlapping area and the template image, , are the offset vectors of the edges of the overlapping and non-overlapping areas, respectively. The edge of the overlapping area , Direction offset distance , constitute, The edge of the non-overlapping area , Direction offset distance , The calculation process of the edge similarity between the overlapping area and the template image and the edge similarity between the non-overlapping area and the template image is consistent with the calculation process of the edge similarity between the closed edge and the template image.

[0015] In the above technical solution, the calculation of the correction distance takes into account multiple key factors. The average color difference between the edge and background color of the overlapping area and the non-overlapping area reflects the color difference information, which is important for distinguishing different areas and evaluating their visual weight in the whole image. Similarity reflects the degree of match between the actual printed edge and the template image edge, which helps to determine the credibility of different areas, while the offset vector represents the deviation of the actual edge from the ideal position, which is the key to calculating the correction direction and distance. Through this comprehensive calculation, the actual deviation of the printed image under the joint influence of the overlapping area and the non-overlapping area can be more accurately reflected.

[0016] Preferably, the method for determining whether there is an overlapping area in the printed images is:

[0017] use Index evaluation Different in clustering The clustering effect of the value When the index is maximum Value as the best value, The value is the number of clusters;

[0018] If the best The value is equal to the number of plates that have been printed plus 1. There is no overlapping area in the printed image. The value is greater than the number of plates that have been printed plus 1, and the printed image has overlapping areas; if the optimal If the value is less than the number of printed plates plus 1, there is no ink at the plate corresponding to the printed image, and an early warning is issued.

[0019] The above technical solution can automatically determine a relatively optimal number of clusters based on the characteristics of image pixels, and can accurately judge whether there are overlapping areas, so that the subsequent processing flow can adopt different processing methods for the two situations where overlapping areas exist and where there are no overlapping areas, thereby improving processing efficiency.

[0020] Preferably, the method for extracting a plurality of edge pixel point sequences of a printed image is:

[0021] Calculate the edge strength of each pixel of the printed image:

[0022] , where is the edge strength of the pixel, The pixel is Brightness gradient in direction, The pixel is Brightness gradient in direction, The pixel is The gradient amplitude of the channel, The pixel is The gradient amplitude of the channel, is a natural constant, It is the symbol for finding the modulus length;

[0023] Based on the edge strength of all pixels and the best Value Clustering is performed to obtain multiple clusters; all the pixels in each cluster except the cluster containing the largest number of pixels are combined into an edge pixel sequence.

[0024] The above technical solution can group the pixels in the image according to the similarity of edge strength through clustering, so that the edge structure of the image is clearer at the level of clustering. Different clusters may correspond to edge areas of different strengths or different types of edge conditions, such as strong edge areas, weak edge areas, etc. This grouping method helps to extract the true edge pixel sequence from the clusters later.

[0025] Preferably, after extracting a plurality of edge pixel point sequences, the method further includes performing the following operations to correct all edge pixel point sequences:

[0026] A set of noise pixels in the printed image is screened out by a filtering algorithm, the set of noise pixels is traversed, and the edge pixels closest to each noise pixel are obtained;

[0027] When the edge strength of a noise pixel is greater than the edge strength of the nearest edge pixel, the distance between the noise pixel and the center of all clusters is further calculated. If the distance between the noise pixel and the center of the cluster where the nearest edge pixel is located is the shortest, the noise pixel is used to replace the nearest edge pixel. If the distance between the noise pixel and the center of the cluster where the nearest edge pixel is located is not the shortest, no replacement is performed, thereby completing the correction of all edge pixel sequences.

[0028] The above technical solution effectively screens out noise pixels through a filtering algorithm, and can separate the interference factors in the image, so that subsequent analysis and processing can focus on these noises that may affect the accuracy of the edge. Obtaining the edge pixels closest to each noise pixel provides a necessary comparison object for further determining whether to replace the edge pixel, which helps to more carefully evaluate the impact of noise on the edge pixel. In addition, this method of determining whether to replace edge pixels based on the dual judgment of the edge strength of the noise pixel and the distance from the center of the cluster can optimize the edge pixel sequence. By replacing the original edge pixels that may not be accurate with noise pixels that are more consistent with edge characteristics, that is, noise pixels with relatively high edge strength and closer to the relevant cluster in spatial position, the accuracy and rationality of the edge pixel sequence can be improved. A more accurate edge pixel sequence helps to more accurately depict the edge shape of the printed image, thereby improving the accuracy of operations related to the edge of the printed image.

[0029] Preferably, before determining whether there is an overlapping area in the printed images, the method further includes:

[0030] Print the image from Color space conversion to Color space, get each pixel Channel value, Channel value and Channel value.

[0031] Preferably, the calculation method of the offset vector of each edge of the overlapping area and the non-overlapping area is:

[0032] A first edge pixel sequence is formed by edge pixel points of the overlapping area / non-overlapping area of ​​the printed image, and a second edge pixel sequence is formed by edge pixel points of the corresponding overlapping area / non-overlapping area of ​​the template image, and linear interpolation is used to make the lengths of the first edge pixel sequence and the second edge pixel sequence the same;

[0033] The average value of the Euclidean distances of edge pixels with the same index in the first edge pixel sequence and the second edge pixel sequence is calculated as an offset vector of the edge of the overlapping area / non-overlapping area of ​​the printed image.

[0034] The above technical solution can accurately locate the area that needs to be compared and analyzed, that is, focus on the edge of the overlapping area and the non-overlapping area. This makes the subsequent calculation and analysis more targeted, and only processes the edge pixels that have a greater impact on the printing quality, avoiding unnecessary complex calculations for the entire image, improving calculation efficiency, and more accurately reflecting the edge deviation that is closely related to the printing quality.

[0035] Preferably, the calculation process of the edge similarity between each closed edge and the template image is:

[0036] Calculate the curvature of each edge pixel of each closed edge, assuming are the 1-step and 3-step curvature change rate sequences of the closed edge, and the total number of curvature change rates is and , They are the 1-step and 3-step curvature change rate sequences of the edge of the template image;

[0037] use The algorithm calculates edge similarity:

[0038] , is the edge similarity between the closed edge and the template image, , yes The mean of yes Middle The sum of the edge strengths of the pixels associated with the curvature change rate, yes Middle The sum of the edge strengths of the pixels associated with the curvature change rate, yes Middle The curvature change rate, yes Middle The curvature change rate, yes and of distance, yes and of distance.

[0039] The above technical solution can more accurately calculate the similarity between the closed edge of the printed image and the edge of the template image. This calculation method that comprehensively considers multiple factors avoids the limitation of relying on a single factor, such as only considering the edge position or only considering the curvature to judge the similarity. For example, even if there are some small deviations in the position of two edges, if their curvature change trends and the edge strength of the relevant pixels are similar, a relatively reasonable similarity measure can still be obtained through this formula. This is very important for accurately evaluating the matching degree between the printed image and the template image in a complex printing environment.

[0040] Preferably, the method for obtaining the brightness gradient and the gradient amplitude is:

[0041] The pixel point and its adjacent pixel point on the right are The difference of the channels is used as the pixel point Brightness gradient in direction ; Place the pixel point and its adjacent pixel point below it in The absolute value of the difference between the channels is used as the pixel point Brightness gradient in direction ;

[0042] The pixel point and its adjacent pixel point on the right are aisle / The absolute value of the difference between the channels is used as the pixel point Gradient magnitude in direction / , the pixel point and its adjacent pixel point below it are aisle / The absolute value of the difference between the channels is used as the pixel point Gradient magnitude in direction / ;

[0043] Pixels in The gradient amplitude of the channel ;

[0044] Pixels in The gradient amplitude of the channel .

[0045] Preferably, filtering out a set of noise pixels in the printed image by a filtering algorithm includes:

[0046] For all pixels The channel values ​​are denoised using wavelet transform to filter out pixels with brightness noise. and The channel uses Gaussian filtering to filter out pixels with color noise, and takes the union of pixels with brightness noise and pixels with color noise as the noise pixel set.

[0047] The present invention has the following effects:

[0048] The present invention classifies and processes by judging whether there is an overlapping area. When there is no overlapping area, the printing plate position is directly matched with the template image by relying on the edge clear feature. When there is an overlapping area, the overlapping area and the non-overlapping area are accurately divided, and a reasonable correction vector is determined according to the offset vector of the edge of each of the overlapping area and the non-overlapping area, which fully and accurately reflects the deviation from the template image, accurately corrects the deviation of the printing plate position, reduces quality problems such as color deviation and pattern misalignment, and improves the matching degree between the printed product and the template image and the overall printing quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0050] Figure 1 It is a schematic flow chart of the method of the present invention;

[0051] Figure 2 The steps of the present invention are Schematic diagram of the method flow chart. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0053] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0054] Reference Figure 1 The method for aligning the position of a color group printing plate provided by the present invention comprises the steps of: -step :

[0055] : Real-time acquisition of the printed image of each printing plate of a multi-color printing plate set.

[0056] The existing gravure printing machine integrates a control and monitoring system. The control system can accurately and automatically record the current printing speed, the number of meters printed during the shift, the cumulative number of meters printed, the strength of the drying box of each printing group, and the tension. (Internet of Things) sensors collect key parameters of the printing press in real time, such as printing speed, drying temperature, operating power, tension, etc. These sensors transmit data to the ERP system through the network to achieve real-time monitoring of equipment status. For example, the smart gateway can access the gravure printing machine (Programmable Logic Controller) collects data and uploads it to the cloud platform or local ERP system. The ERP system monitors the production process in real time based on the received equipment data. Once the data is abnormal, such as excessive temperature, abnormal speed, etc., the system will automatically trigger an alarm and notify relevant personnel to handle it in time.

[0057] In order to more accurately ensure the printing quality, this step installs a camera at each printing plate to collect the printing image at each printing plate, and transmits the printing image collected by the camera, the print quality screen, and the real-time print image and template image at the printing plate to the ERP system through the network protocol (such as TCP / IP protocol). The ERP system identifies and receives various data and images transmitted from the printing press, and classifies and stores the received data according to the data type and source. The printing press operation data is stored in a specific table in a relational database (such as MySQL or Oracle), and each data item corresponds to a field in the table; the image data is stored in the file system, and a directory structure is established according to the shooting time, printing plate number and other information for storage.

[0058] The ERP system has a built-in data display function, which is used to view all the working status data of the printing machine in real time, including display in the form of charts (such as bar charts showing printing speed changes, line charts showing power fluctuations, etc.), tables (such as data lists of printed meters, drying box strength, tension size, etc.). The ERP system has a built-in image viewing function for viewing printed images. It can retrieve and view the real-time working screen, print quality screen, and real-time print image and template image at the printing plate according to conditions such as time and printing plate number. At the same time, it provides an image magnification function, especially for the print quality screen, which can view the print quality in detail. The ERP system has built-in data analysis functions, such as calculating the average printing speed in different time periods, power consumption trend analysis, and yield rate change analysis, etc., to provide data support for production management, implementation of production plans, and production cost accounting. By analyzing the data, abnormal conditions in the printing process (such as a sudden drop in printing speed, abnormal increase in power, and reduced yield rate) can be discovered in time, and corresponding measures can be taken to adjust and optimize. Through the above complete steps, various key data and images of the gravure printing machine can be transmitted to the real-time system in real time. In the system, comprehensive grasp and effective management of the working status of the printing press can be achieved.

[0059] It should be noted that the current printing speed refers to the rate at which the printing material (such as paper) passes through the printing unit when the printing machine is printing at a certain moment. It is usually expressed in units such as meters per minute or feet per minute. The printing machine measures this data through speed sensors installed in the transmission system. These sensors can be encoders, which are connected to the drive shaft or guide roller of the printing machine. As the machine runs, the encoder calculates the printing speed based on the rotation angle and speed of the shaft. The number of printed meters on a shift refers to the total length of materials printed by the printing machine in a shift (usually a working period, such as 8 hours). It is calculated by integrating the printing speed over time. The control system of the printing machine will record the current printing speed once at a certain interval (such as every minute), multiply these speed values ​​by the corresponding time interval, and then add them up to obtain the number of printed meters on a shift. In addition, it can also be directly measured by installing a meter meter on the printing material conveying path. The cumulative printing meters refers to the total length of the material printed by the printing machine from the beginning of use or a certain set initial time point. Its measurement principle is similar to the printing meters on duty, except that the printing speed is integrated over a long period of time during the entire use cycle. The control system of the printing machine continuously accumulates the length data of each printing job and stores it in a non-volatile memory to obtain the cumulative printing meters. The strength of the drying box of each printing group usually refers to the comprehensive index of parameters such as heating power, hot air speed, and temperature in the drying box. The drying box is equipped with temperature sensors, wind speed sensors, etc., which transmit the collected data to the control system. The tension refers to the pulling force on the printing material in the longitudinal direction (consistent with the printing direction) and the transverse direction (perpendicular to the printing direction) during the conveying process of the printing machine. The printing machine measures the tension through the tension sensors installed on the material conveying path. These sensors can be strain gauge sensors or other types of force sensors, which convert the tension into electrical signals and then transmit them to the control system. The control system maintains appropriate tension by adjusting the tension control system (such as the position of the tension roller, the torque of the motor, etc.) according to the type of printing material (such as paper thickness, elasticity, etc.), printing speed and printing process requirements.

[0060] : Determine whether there is an overlapping area in the printed image.

[0061] In step The printed image at each printing plate has been collected. For any printed image at a printing plate, due to environmental noise such as lighting environment and paper vibration, pseudo edge points will be introduced, causing the edge detection algorithm to misjudge, resulting in edge distortion when extracting edges. At the same time, the color brightness is different after different colors are mixed and overlapped. The grayscale image denoising converted from RGB images will have problems such as too high signal-to-noise ratio or low noise removal efficiency. Therefore, it is necessary to denoise the real-time printed image. The specific method is: first convert the real-time printed image from RGB to LAB color space. "L" in LAB color space represents brightness, and its value range is 0-100, 0 represents black, and 100 represents white; "A" represents the color change axis from green to red, and the value range is generally -128 to 127, negative numbers represent the green direction, and positive numbers represent the red direction; "B" represents the color change axis from blue to yellow, and the value range is generally -128 to 127, negative numbers represent the blue direction, and positive numbers represent the yellow direction. Then, wavelet transform is used to denoise the L channel of the LAB image to filter out brightness noise pixels, and Gaussian filtering is used to denoise the A and B channels to filter out color noise pixels. The union of brightness noise pixels and color noise pixels is taken, and the union is used as the noise pixel set. Then, the LAB image is converted into a grayscale image. Therefore, the printed images referred to below are all grayscale images.

[0062] The method for determining whether there is an overlapping area in the printed images is as follows:

[0063] use index( Index) Assessment Different in clustering The clustering effect of the value When the index is maximum Value as the best value, The value is the number of clusters.

[0064] If the best The value is equal to the number of plates that have been printed plus 1. There is no overlapping area in the printed image. The value is greater than the number of plates that have been printed plus 1, and the printed image has overlapping areas; if the optimal If the value is less than the number of printed plates plus 1, there is no ink at the plate corresponding to the printed image, and an early warning is issued.

[0065] It should be noted that The index is an indicator used to evaluate clustering effects. In the analysis of printed images, different pixel values ​​in the image are regarded as different data points for clustering. This index measures the quality of clustering based on the compactness within the cluster and the separation between clusters. Specifically, its calculation involves the variance within the cluster and the variance between clusters. When the pixel values ​​within the cluster are relatively close and the distance between different clusters is large, the index will be larger, indicating that the clustering effect is better.

[0066] It should be noted that during the printing process, each printing plate will theoretically form an independent area on the image. If the printed image has no overlapping areas, the clustering of pixel values ​​should be relatively clear, that is, the number of clusters should be equal to the number of printed plates plus 1 (taking into account factors such as background). For example, assuming that there are two printing plates for printing, and there is no overlap under normal circumstances, then the pixel values ​​of the image should be clearly divided into 3 clusters: the areas corresponding to the two printing plates and the background. At this time, the best The value is equal to the number of plates that have been printed plus 1. The value is 3. If there are overlapping areas in the printed images, the originally clear pixel value clustering will become complicated, and the pixel values ​​originally belonging to different printing plates will be mixed in the overlapping parts, which will increase the number of clusters. Therefore, if the number of clusters is greater than the number of printed plates plus 1, it can be judged that there is an overlapping area. If the best The value is less than the number of printed plates plus 1, which means that the pixel values ​​actually involved in forming effective clusters are reduced. From the perspective of printing, it is likely that there is no ink in the area corresponding to a printing plate, resulting in the failure of the clusters that should have been formed to form normally, thereby reducing the number of clusters. This situation is like a missing printing plate leaving a mark on the image, so it can be judged that there is no ink at the printing plate corresponding to the printed image.

[0067] Next, if there is no overlapping area in the printed images, execute step If there are overlapping areas in the printed images, execute step .

[0068] : Match the printed image directly with the template image to align the position of the printing plate corresponding to the printed image.

[0069] Here, a matching algorithm based on feature points is used (e.g. , etc.), by finding the corresponding relationship of feature points or similar areas between the printed image and the template image, the relative position deviation between the printed image and the template image is determined, and the position of the printing plate corresponding to the printed image is aligned according to the deviation to ensure the accuracy of the printing plate position.

[0070] :Accurately divide the overlapping area and the non-overlapping area, and determine the correction vector of the printed image according to the offset of the overlapping area and the non-overlapping area, and use the correction vector to align the position of the printing plate corresponding to the printed image.

[0071] Specific reference Figure 2 , including the following steps:

[0072] : Calculate the edge strength of each pixel based on the brightness gradient and color gradient, using The clustering algorithm is used to cluster and extract multiple edge pixel sequences of the printed image.

[0073] First, calculate the edge strength of each pixel of the printed image:

[0074]

[0075] In the formula, is the edge strength of the pixel, The pixel is Brightness gradient in direction, The pixel is Brightness gradient in direction, The pixel is The gradient amplitude of the channel, The pixel is The gradient amplitude of the channel, is a natural constant, It is the symbol for finding the modulus length;

[0076] In this formula, the method for obtaining the brightness gradient and gradient amplitude is:

[0077] The pixel point and its adjacent pixel point on the right are The difference of the channels is used as the pixel point Brightness gradient in direction ; Place the pixel point and its adjacent pixel point below it in The absolute value of the difference between the channels is used as the pixel point Brightness gradient in direction ;

[0078] The pixel point and its adjacent pixel point on the right are The absolute value of the difference between the channels is used as the pixel point Gradient magnitude in direction , the pixel point and its adjacent pixel point below it are The absolute value of the difference between the channels is used as the pixel point Gradient magnitude in direction , then the pixel is The gradient amplitude of the channel .

[0079] The pixel point and its adjacent pixel point on the right are The absolute value of the difference between the channels is used as the pixel point Gradient magnitude in direction , the pixel point and its adjacent pixel point below it are The absolute value of the difference between the channels is used as the pixel point Gradient magnitude in direction , then the pixel is The gradient amplitude of the channel .

[0080] Based on the edge strength of all pixels and the best Value Clustering is performed to obtain multiple clusters, and the pixels of each cluster other than the cluster with the largest number of pixels are regarded as an edge pixel sequence. This processing method is based on the fact that printed images have relatively few edge pixels. In an image, the edge part is usually the boundary between different objects (in a printed image, it can be different printing plate patterns, etc.), and the number of pixels is often less than that of the non-edge area. In this way, these edge pixels can be accurately extracted. This method avoids the problem of misjudging a large number of non-edge pixels as edge pixels, because the largest cluster usually corresponds to the non-edge area in the image (such as a large area of ​​solid color background or the inside of the main printed pattern).

[0081] In one embodiment, multiple edge pixel point sequences are corrected according to the noise pixel point set, and the specific correction process is:

[0082] The noise pixel set is traversed, and the nearest edge pixel of each noise pixel is obtained. When the edge strength of a noise pixel is greater than the edge strength of the nearest edge pixel, the distance between the noise pixel and the center of all clusters is further calculated. If the distance between the noise pixel and the center of the cluster where the nearest edge pixel is located is the shortest, the nearest edge pixel is replaced by the noise pixel. If the distance between the noise pixel and the center of the cluster where the nearest edge pixel is located is not the shortest, no replacement is performed, thereby completing the correction of all edge pixel sequences.

[0083] It should be noted that noise is inevitable in the process of printing image processing. These noise pixels may be caused by factors such as slight jitter of the printing equipment, unevenness of the paper, granularity of the ink, or interference of electronic equipment during the image acquisition process. The presence of noise pixels will interfere with the accuracy of the edge pixel sequence because they may be mistakenly identified as edge pixels or cause the real edge pixels to be masked. For example, if an isolated noise pixel has a high edge intensity value and is near an edge pixel, the shape and position of the edge may be misjudged.

[0084] An accurate edge pixel sequence is crucial for subsequent printing image processing. For example, in operations such as plate position correction, image matching, and printing quality assessment, the edge pixel sequence is a key reference. If the edge pixel sequence is interfered with by noise, then operations based on this edge information (such as calculating the correction distance, determining the overlapping area of ​​the image, etc.) will produce errors, which will affect the final quality of the printed product. Therefore, in order to ensure that the edge pixel sequence can truly reflect the edge information of the printed image, it needs to be corrected to eliminate the interference of noise pixels.

[0085] When the edge strength of a noise pixel is greater than the edge strength of the nearest edge pixel, further operation is considered. This is because if the edge strength of the noise pixel is low, then its impact on the edge pixel sequence is relatively small and can be ignored. Only when the noise pixel has a high enough edge strength and is likely to change the shape or position of the edge, is it necessary to make subsequent judgments and processing.

[0086] Calculate the distance between the noise pixel and the center of all clusters, and decide whether to replace it according to the distance to the center of the cluster where the nearest edge pixel is located. This is a method that comprehensively considers the spatial position relationship of pixels. If the noise pixel is at the shortest distance from the center of the cluster where the nearest edge pixel is located, it means that it is more inclined to belong to this cluster in terms of spatial position. At this time, replacing the nearest edge pixel with the noise pixel can make the edge pixel sequence more reasonable in spatial distribution. On the contrary, if the distance is not the shortest, it means that the noise pixel is more inclined to other clusters in space. Not replacing it can avoid introducing erroneous pixels into the edge pixel sequence, thereby maintaining the accuracy of the edge pixel sequence. In this way, the effective edge information that may be contained in the noise pixel can be utilized to a certain extent, while effectively eliminating the interference of noise on the edge pixel sequence.

[0087] : Get the closed edge and determine the actual edge of the printed image based on the similarity between the closed edge and the edge of the template image.

[0088] First, any two edge pixel sequences that can form a closed figure are regarded as a closed edge, and multiple closed edges in the printed image can be obtained. Then, the edge similarity between each closed edge and the template image is calculated as follows:

[0089] Calculate the curvature of each edge pixel of each closed edge, and specifically use the curvature calculation formula based on difference. The specific method is:

[0090] Get the coordinates of each edge pixel of the closed edge. edge pixels, whose coordinates for , Indicates Pixels coordinate, Indicates Pixels coordinate.

[0091] Calculate the pixel point First-order difference of direction , the pixel is at First-order difference of direction ,in, Indicates Pixels coordinate, Indicates Pixels coordinate.

[0092] Next, calculate the pixel point Second-order difference of direction , calculate the pixel point at Second-order difference of direction ,in, Indicates Pixels coordinate, Indicates Pixels coordinate.

[0093] Finally, the curvature calculation formula based on difference is obtained:

[0094]

[0095] In the formula, For the The curvature of the edge pixels.

[0096] Different curvature change rate sequences of each closed edge are obtained according to different step sizes. When the step size is 1, that is, each edge pixel of the closed edge is traversed one by one, a curvature change rate is obtained for every two adjacent edge pixels, and a 1-step curvature change rate sequence is obtained. When the step size is 3, a curvature change rate is calculated for every 3 edge pixels, and a 3-step curvature change rate sequence is obtained.

[0097] set up and are the 1-step curvature change rate sequence and the 3-step curvature change rate sequence of the closed edge respectively. ， and They are the 1-step curvature change rate sequence and the 3-step curvature change rate sequence of the edge of the template image respectively;

[0098] use The (Dynamic Time Warping) algorithm calculates the edge similarity between each closed edge and the template image:

[0099]

[0100] In the formula, is the edge similarity between the closed edge and the template image, yes The mean of for The mean of yes Middle The sum of the edge strengths of the pixels associated with the curvature change rate, yes Middle The sum of the edge strengths of the pixels associated with the curvature change rate, yes Middle The curvature change rate, yes Middle The curvature change rate, yes and of distance, yes and of distance, is the total number of curvature change rates contained in the 1-step curvature change rate sequence of the closed edge, is the total number of curvature change rates contained in the 3-step curvature change rate sequence of the closed edge.

[0101] In this formula, the total number of curvature change rates contained in the curvature change rate sequence of the closed edge (1-step curvature change rate sequence and 3-step curvature change rate sequence) is ( and ) is used as the standard, a corresponding number of curvature change rate sequences are obtained in the edge of the template image, and then the similarity between the two is calculated.

[0102] In this formula, calculating the 1-step curvature change rate sequence and the 3-step curvature change rate sequence of the closed edge and the template image edge is the key to measuring the edge shape change. In a printed image, the curvature change of the edge can well reflect the shape characteristics of the pattern. For example, for the edge of a circular pattern, its curvature change is relatively stable, while for a pattern with sharp corners, the curvature change will fluctuate greatly at the corner. By comparing the curvature change rate of the closed edge of the printed image and the edge of the template image, the degree of similarity in their shapes can be evaluated.

[0103] This formula makes the calculation of similarity more comprehensive by introducing edge strength combined with the curvature change rate. Edge strength reflects the prominence of edge pixels. In printed images, clear and distinct edges usually have higher edge strength. Multiplying the sum of edge strengths by the average curvature change rate means that when evaluating similarity, not only the shape change of the edge is considered, but also the prominence of the edge. This helps to more accurately determine whether the edge of the printed image is truly similar to the edge of the template image, because even if the shape is similar, if the edge strength difference is large, it may mean that there is a printing quality problem or the edge is not truly matched.

[0104] In this formula and The method to obtain is:

[0105] because The (Dynamic Time Warping) algorithm is an algorithm used to measure the similarity of two time series (here, the curvature change rate series). Its basic idea is to find the optimal alignment between the two sequences so that their differences on the time axis are minimized. In the comparison of the curvature change rate series of the edges of printed images, due to various factors in the printing process, such as stretching and deformation, the curvature change rate series of the two edges may not be completely aligned in time (or the order of edge pixels).

[0106] For two curvature change rate sequences, such as and , The algorithm first constructs a cost matrix, the elements of which represent the cost required to align a point in one sequence with a point in another sequence (usually a measure of the difference between the two points). Then, through dynamic programming, starting from the upper left corner of the matrix, find a path from the upper left corner to the lower right corner so that the sum of the costs on the path is minimized. This path represents the optimal alignment of the two sequences, and the minimum sum of costs is Distance, obtained in this way and .

[0107] Next, the closed edge with the greatest similarity to the edge of the template image is taken as the actual edge of the printed image, and each printed image will obtain an actual edge.

[0108] :Divide the overlapping area and non-overlapping area according to the actual edge of the printed image.

[0109] Clustering is performed on the area within the actual edge of the printed image, where A clustering algorithm is used to obtain multiple clusters, each cluster contains multiple pixel points. Then, the sum of the distances from all pixel points of each cluster to the center points of all other clusters is calculated, and the cluster with the smallest sum of distances is taken as the overlapping area, and the area in the printed image except the overlapping area is taken as the non-overlapping area.

[0110] It should be noted that in the printed image, the shape of the area formed by the printing plate may be irregular. The density-based properties of the Iterative Self-Organizing Clustering (ITSC) algorithm enable it to discover clusters of arbitrary shapes. In a printed image, the overlapping area is the part where multiple plate patterns intersect each other. From the perspective of the spatial distribution of pixels, the pixels in the overlapping area are closely related to the non-overlapping areas corresponding to multiple different plates. It is at the intersection of different plate patterns, and its pixel distribution is spatially closer to the center of other areas. For example, suppose there are two plates. and , the pixels in the overlapping area include both the printing plate Some pixel features of the printing plate The characteristics of some pixels. This makes the sum of the overall distances of these pixels relatively small when calculating the distance to the center of other clusters, because they tend to be distributed in the "middle" position of different non-overlapping areas in space. Calculate the sum of the distances from all pixels of each cluster to the center points of all other clusters. This distance calculation method actually measures the spatial correlation between one cluster and other clusters. Since the overlapping area is the overlap of multiple printing plate patterns, its spatial correlation with other areas is the most complex. The non-overlapping area is mainly related to the printing plate pattern corresponding to itself, and the distance from its pixel to the center of other clusters is relatively large. Because the overlapping area has to consider the influence of multiple printing plate patterns at the same time, its pixel points are closer to the center of other clusters in spatial distribution, so it will show the minimum value when calculating the sum of distances, so it can be identified as an overlapping area.

[0111] : Calculate the offset vectors of the edges of the overlapping area and the non-overlapping area, and then obtain the correction vector of the printed image.

[0112] First, the offset vectors of the edges of the overlapping and non-overlapping regions are calculated as follows:

[0113] For the offset vector of the edge of the overlapping area: the edge pixels of the overlapping area of ​​the printed image constitute the first edge pixel sequence, and the edge pixels of the corresponding overlapping area of ​​the template image constitute the second edge pixel sequence. Through linear interpolation, the lengths of the first edge pixel sequence and the second edge pixel sequence are made the same. The mean of the Euclidean distances of the edge pixels with the same index in the first edge pixel sequence and the second edge pixel sequence is calculated as the offset vector of the edge of the overlapping area of ​​the printed image, and used indicates that, and , Indicates that the edge of the overlapping area is The offset distance of the direction, Indicates that the edge of the overlapping area is The offset distance of the direction.

[0114] The offset vector of the edge of the overlapping area is obtained in the same way as the overlapping area, that is:

[0115] The edge pixels of the non-overlapping area are used to form a first edge pixel sequence, and the edge pixels of the corresponding non-overlapping area of ​​the template image are used to form a second edge pixel sequence. The lengths of the first edge pixel sequence and the second edge pixel sequence are made the same through linear interpolation. The mean of the Euclidean distances of the edge pixels with the same index in the first edge pixel sequence and the second edge pixel sequence is calculated as the offset vector of the edge of the non-overlapping area of ​​the printed image. indicates that, and , The edge of the non-overlapping area is The offset distance of the direction, The edge of the non-overlapping area is The offset distance of the direction.

[0116] Next, the edge similarity between the overlapping area and the non-overlapping area and the edge similarity between the template image is analyzed. Since the edges of the overlapping area and the non-overlapping area are also closed edges, the edge similarity between the edges of the overlapping area and the non-overlapping area and the template image is calculated according to the method of calculating the edge similarity between the closed edge and the template image in step S222.

[0117] Finally, calculate the correction vector of the printed image according to the following formula:

[0118]

[0119] In the formula, is the correction vector for the printed image, The printed images are , Correction distance for direction , Composition, that is , , are the average color difference between the edge and background colors of the overlapping area and the non-overlapping area, , They are the similarities between the edges of the overlapping area and the non-overlapping area and the edge of the template image. , They are the offset vectors of the edges of the overlapping area and the non-overlapping area respectively.

[0120] In this formula, , It reflects the degree of color difference between the edges and backgrounds of different regions. The greater the color difference, the greater the weight of a region in the position correction process. , It reflects the similarity between the edges of different regions and the edges of the template image. The higher the similarity, the better the match between the region and the template, and the greater the contribution to the calculation of the correction distance.

[0121] It should be noted that the calculation process of the edge similarity between the overlapping area of ​​the printed image and the template image, and the edge similarity between the non-overlapping area and the template image are consistent with the calculation process of the edge similarity between the closed edge and the template image.

[0122] : Based on the correction vector, the printing plate corresponding to the printed image is aligned.

[0123] Getting the corrected distance After that, the position of the printing plate corresponding to the printed image can be corrected. For the gravure printing machine, the printing plate is installed on a device that can be moved accurately. The control system will issue instructions to move the printing plate according to the calculated correction distance. The distance moved in the direction is ,exist The distance moved in the direction is , so that the position of the printed image is accurately aligned with the position of the template image, ensuring the accuracy of subsequent printing.

[0124] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0125] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for aligning a color group printing plate, characterized in that: include: Acquire the printed image of each printing plate of the multi-color printing plate set in real time, determine whether there is an overlapping area of ​​the printed image, and if there is no overlapping area, directly match the printed image with the template image to align the position of the printing plate corresponding to the printed image; if there is an overlapping area, perform the following operations: Extract multiple edge pixel sequences of the printed image, take any two edge pixel sequences that can form a closed figure as a closed edge, calculate the edge similarity between each closed edge and the template image, and take the closed edge with the largest edge similarity as the actual edge of the printed image; the calculation process of the edge similarity between each closed edge and the template image is: Calculate the curvature of each edge pixel of each closed edge, assuming are the 1-step and 3-step curvature change rate sequences of the closed edge, and the total number of curvature change rates is and , They are the 1-step and 3-step curvature change rate sequences of the edge of the template image; use The algorithm calculates edge similarity: , is the edge similarity between the closed edge and the template image, , yes The mean of yes Middle The sum of the edge strengths of the pixels associated with the curvature change rate, yes Middle The sum of the edge strengths of the pixels associated with the curvature change rate, yes Middle The curvature change rate, yes Middle The curvature change rate, yes and of distance, yes and of distance; Clustering the area within the actual edge of the printed image to obtain multiple clusters, and calculating the sum of distances from all pixel points of each cluster to the center points of all other clusters, taking the cluster with the smallest sum of distances as the overlapping area, and taking the area of ​​the printed image other than the overlapping area as the non-overlapping area; Calculating the offset vectors of the edges of the overlapping area and the non-overlapping area in combination with the edge of the template image, determining the correction vector of the printed image based on the offset vectors of the edges of the overlapping area and the non-overlapping area, and aligning the printing plate corresponding to the printed image based on the correction vector; The calculation method of the offset vector of the edge of the overlapping area and the non-overlapping area is: A first edge pixel sequence is formed by edge pixel points of the overlapping area / non-overlapping area of ​​the printed image, and a second edge pixel sequence is formed by edge pixel points of the corresponding overlapping area / non-overlapping area of ​​the template image, and the lengths of the first edge pixel sequence and the second edge pixel sequence are made the same through linear interpolation; the mean of the Euclidean distances of edge pixel points with the same index in the first edge pixel sequence and the second edge pixel sequence is calculated as the offset vector of the edge of the overlapping area / non-overlapping area of ​​the printed image; The correction vector of the printed image satisfies the following relationship: , where is the correction vector for the printed image, The printed images are , Correction distance for direction , constitute, , are the average color difference between the edge and background colors of the overlapping area and the non-overlapping area, , They are the edge similarities between the overlapping area, the non-overlapping area and the template image, , are the offset vectors of the edges of the overlapping and non-overlapping areas, respectively. The edge of the overlapping area , Direction offset distance , constitute, The edge of the non-overlapping area , Direction offset distance , The calculation process of the edge similarity between the overlapping area and the template image and the edge similarity between the non-overlapping area and the template image is consistent with the calculation process of the edge similarity between the closed edge and the template image.

2. The method for aligning the position of a color group printing plate according to claim 1, characterized in that: The method for determining whether there is an overlapping area in the printed image is: use Index evaluation Different in clustering The clustering effect of the value When the index is maximum Value as the best value, The value is the number of clusters; If the best The value is equal to the number of plates that have been printed plus 1. There is no overlapping area in the printed image. The value is greater than the number of plates that have been printed plus 1, and the printed image has overlapping areas; if the optimal If the value is less than the number of printed plates plus 1, there is no ink at the plate corresponding to the printed image, and an early warning is issued.

3. The method for aligning the position of a color group printing plate according to claim 1, characterized in that: The method for extracting multiple edge pixel sequences of a printed image is: Calculate the edge strength of each pixel of the printed image: , where is the edge strength of the pixel, The pixel is Brightness gradient in direction, The pixel is Brightness gradient in direction, The pixel is The gradient amplitude of the channel, The pixel is The gradient amplitude of the channel, is a natural constant, It is the symbol for finding the modulus length; Based on the edge strength of all pixels and the best Value Clustering is performed to obtain multiple clusters; all the pixels in each cluster except the cluster containing the largest number of pixels are combined into an edge pixel sequence.

4. The method for aligning the position of a color group printing plate according to claim 1, characterized in that: After extracting multiple edge pixel point sequences, the method further includes performing the following operations to correct all edge pixel point sequences: A set of noise pixels in the printed image is screened out by a filtering algorithm, the set of noise pixels is traversed, and the edge pixels closest to each noise pixel are obtained; When the edge strength of a noise pixel is greater than the edge strength of the nearest edge pixel, the distance between the noise pixel and the center of all clusters is further calculated. If the distance between the noise pixel and the center of the cluster where the nearest edge pixel is located is the shortest, the noise pixel is used to replace the nearest edge pixel. If the distance between the noise pixel and the center of the cluster where the nearest edge pixel is located is not the shortest, no replacement is performed, thereby completing the correction of all edge pixel sequences.

5. The method for aligning the position of a color group printing plate according to claim 1, characterized in that: Before determining whether there is an overlapping area in the printed images, the method further includes: Print the image from Color space conversion to Color space, get each pixel Channel value, Channel value and Channel value.

6. The method for aligning the position of a color group printing plate according to claim 3, characterized in that: The method for obtaining the brightness gradient and gradient amplitude is: The pixel point and its adjacent pixel point on the right are The difference of the channels is used as the pixel point Brightness gradient in direction ; Place the pixel point and its adjacent pixel point below it in The absolute value of the difference between the channels is used as the pixel point Brightness gradient in direction ; The pixel point and its adjacent pixel point on the right are The absolute value of the difference between the channels is used as the pixel point Channel Gradient magnitude in direction , the pixel point and its adjacent pixel point below it are The absolute value of the difference between the channels is used as the pixel point Channel Gradient magnitude in direction ; The pixel point and its adjacent pixel point on the right are The absolute value of the difference between the channels is used as the pixel point Channel Gradient magnitude in direction , the pixel point and its adjacent pixel point below it are The absolute value of the difference between the channels is used as the pixel point Channel Gradient magnitude in direction ; Pixels in The gradient amplitude of the channel ; Pixels in The gradient amplitude of the channel .

7. The method for aligning the position of a color group printing plate according to claim 4, characterized in that: The noise pixel set in the printed image is filtered out through the filtering algorithm, including: For all pixels The channel values ​​are denoised using wavelet transform to filter out pixels with brightness noise. and The channel uses Gaussian filtering to filter out pixels with color noise, and takes the union of pixels with brightness noise and pixels with color noise as the noise pixel set.

Images