Image-text defect detection method and system for color synthesis based on plate roller image

Abstract

The invention discloses an image-text defect detection method and system for color synthesis based on a plate roller image, and belongs to the technical field of printing quality detection. The method comprises the steps that a plate roller image comprising screen dots is obtained, an actual screen dot model is established, and the plate roller image is obtained by conducting three-dimensional scanning on a plate roller; converting a pre-acquired manuscript image into a theoretical dot model, and identifying dot defects based on the theoretical dot model and an actual dot model; predicting a printing effect image after color synthesis based on the plate roller image; color defects are identified based on the printing effect image and the document image. The method further comprises a step of calculating saturation based on the dot volume and performing color synthesis, and a step of sorting dot defects by comparing model differences, semantic segmentation and object-severity rules. The printing effect can be predicted before printing, dot defects and color defects can be detected, and the printing quality control efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of printing quality detection, and in particular to a method and system for detecting image defects by performing color synthesis based on a plate roller image. Background Art

[0002] With the continuous advancement of printing technology, the importance of printed product quality inspection has become increasingly prominent. In the gravure printing process, the plate roller is a key printing component, and the quality of the dots on its surface directly affects the final printed product. Currently, the gravure printing industry relies primarily on proofing and manual visual inspection to inspect plate roller quality, which is inefficient and highly subjective.

[0003] Numerous methods and systems exist for detecting defects in printed products. For example, Chinese patent application publication number CN119417779A discloses an intelligent defect recognition system for sublimation transfer paper based on image analysis. This system includes an image acquisition module, a pattern defect analysis module, and a color defect analysis module. By analyzing the pattern and color defect parameters of the transferred finished image, it generates corresponding evaluation coefficients, which are then comprehensively evaluated to determine the defect evaluation index for the sublimation transfer paper. While this system can detect defects in printed products, it primarily targets the final printed product, not the pre-printing plate roller, and cannot detect pre-printing problems with the plate roller.

[0004] Chinese patent application publication number CN112541889A proposes a multi-layered model for detecting surface defects on complex textured objects. This method achieves real-time detection of surface defects on complex textured objects by hierarchically connecting three models: a detection rule-based model, a pattern recognition-based model, and a deep learning model. However, this method primarily focuses on detecting defects on general textured surfaces and does not consider the unique three-dimensional structural characteristics of plate roller dots, nor does it address the technical means of color synthesis to predict printing effects.

[0005] Regarding color defect detection, Chinese Patent Application Publication No. CN114549418A discloses a defect detection method based on a scoring network. This method constructs feature vectors based on multiple color spaces and quantitatively scores each pixel in a photograph of a wood panel to identify defects. Chinese Patent Application Publication No. CN117372331A proposes a method for detecting color defects on printed surfaces. This method obtains a frequency-domain difference image between a target image and a defect-free image, performs a series of processing operations, and then determines whether a defect exists. While these methods offer some innovations in color defect detection, none of them directly correlates the three-dimensional characteristics of the plate roller dots with the final print result, making it impossible to predict potential color issues before printing. Chinese Patent Application Publication No. CN113888533A describes a display panel defect detection device. This device determines defect coordinates by capturing a feature image, then captures a local color image and uses a defect recognition model for classification. While this method achieves accurate defect classification, it also fails to consider the relationship between the unique three-dimensional structure of the plate roller and the print result.

[0006] The existing plate roller inspection technology has the following main problems: First, the traditional proofing inspection method requires trial printing or proofing, which is not only unprofitable but also produces unnecessary VOC emissions and causes environmental pollution; second, the existing inspection method is highly dependent on personnel, and the inspectors may miss inspections if they are not careful. Once the defective plate roller enters the mass production link, it will cause serious economic losses; third, the proofing and inspection processes require manual operation, which is inefficient and difficult to automate; finally, the existing technology fails to establish a direct correlation between the three-dimensional characteristics of the plate roller dots and the final printing effect, and cannot accurately predict possible color problems such as overprinting and moiré before printing.

[0007] Therefore, there is an urgent need for a method that can perform defect detection directly based on the plate roller image before printing, which can not only detect dot defects but also predict the printing effect through color synthesis, so as to discover and solve possible problems in advance and improve the automation level and efficiency of printing quality control. Summary of the Invention

[0008] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method and system for detecting image defects based on color synthesis of plate roller images, so as to solve or partially solve the problem that the existing gravure plate inspection method is prone to miss detection of plate roller defects through proofing and the degree of automation is not ideal. The existing gravure plate inspection method is to inspect the sample printed by trial printing, which has the following defects: (1) trial printing or proofing does not produce benefits and will produce unnecessary VOCs emissions; (2) the existing inspection method is highly dependent on human labor, and even the slightest carelessness will result in missed detection, which will cause economic losses when entering mass production; (3) proofing and inspection are both inseparable from human labor, low efficiency and cannot be automated.

[0009] The purpose of the present invention can be achieved by the following technical solutions:

[0010] One aspect of the present invention provides a method for detecting image defects based on color synthesis of a plate roller image, comprising the following steps:

[0011] Step S1, obtaining an image of a plate roller including halftone dots and establishing an actual halftone dot model, wherein the image of the plate roller is obtained by performing a three-dimensional scanning on the plate roller;

[0012] Step S2, converting the pre-acquired original image into a theoretical dot model, and identifying dot defects based on the theoretical dot model and the actual dot model;

[0013] Step S3, predicting a printing effect image after color synthesis based on the plate roller image;

[0014] Step S4: identifying color defects based on the printing effect image and the original image.

[0015] As a preferred technical solution, step S3 includes the following sub-steps:

[0016] Step S301, calculating the volume of each dot based on the plate roller image;

[0017] Step S302, converting the volume of each dot into the saturation of each dot;

[0018] Step S303 : obtaining a printing effect image by color synthesis based on the saturation of each dot and the pre-acquired ink-carrying color of each dot.

[0019] As a preferred technical solution, step S2 includes the following sub-steps:

[0020] Step S201, obtaining the position of each dot defect by comparing the difference between the theoretical dot model and the actual dot model;

[0021] Step S202 , performing semantic segmentation on the original image, and obtaining object information where each dot defect is located based on the location of each dot defect;

[0022] Step S203 : sorting the dot defects based on the object information of each dot defect and the pre-built object-severity rules.

[0023] As a preferred technical solution, it also includes:

[0024] Step S5: superimpose the color defect and the halftone dot defect on the plate roller image, highlight the position corresponding to the defect, and output it to a visualization terminal.

[0025] As a preferred technical solution, the color defects include overprinting and moire.

[0026] Another aspect of the present invention provides a graphic defect detection system for color synthesis based on a plate roller image, comprising:

[0027] Plate roller fixing module, used to fix the plate roller to be tested;

[0028] A three-dimensional scanning module is used to perform three-dimensional scanning on the fixed plate roller to obtain an image of the plate roller including dots;

[0029] The data processing module is used to realize defect detection based on the plate roller image by the above-mentioned image and text defect detection method of color synthesis based on the plate roller image.

[0030] As a preferred technical solution, it also includes:

[0031] The database stores object-severity rules for ranking dot defects.

[0032] As a preferred technical solution, it also includes:

[0033] Visualization terminal, used to display the results of defect identification.

[0034] Another aspect of the present invention provides an electronic device comprising: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the aforementioned graphic defect detection method based on color synthesis of a plate roller image.

[0035] Another aspect of the present invention provides a computer-readable storage medium comprising one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the aforementioned method for detecting graphic defects by color synthesis based on a plate roller image.

[0036] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0037] (1) Automatic detection of image and text defects is achieved: the present invention establishes an actual dot model by acquiring the plate roller image, and converts the original image into a theoretical dot model. Dot defects are identified by comparing the two models. At the same time, the printing effect image is predicted based on the plate roller image, and color defects are identified by comparing it with the original image. There is no need for actual proofing, which avoids unnecessary VOCs emissions, reduces manual intervention, and improves detection efficiency.

[0038] (2) Capable of detecting various types of defects: On the one hand, the present invention can detect defects in the dots themselves through model comparison; on the other hand, it can predict the printing effect diagram and identify color defects by comparing with the original image. It can detect plate roller defects more comprehensively, ensure the accuracy of the detection results, and effectively avoid the problem of missed detection in manual detection.

[0039] (3) More intuitive display of defects: The present invention solves the problem of being unable to intuitively discover key defects during the inspection process by sorting the dot defects and superimposing the color defects and dot defects on the plate roller image for highlighting. At the same time, the system outputs the defects to the visual terminal, making it easier for operators to quickly locate the defect position, thereby improving the detection efficiency and providing technical support for the realization of automated detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Flowchart of a method for detecting image defects based on color synthesis of a plate roller image in an embodiment;

[0041] Figure 2 Schematic diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0043] Example 1

[0044] Regarding the problems existing in the above-mentioned prior art, see Figure 1 This embodiment provides a method for detecting image defects based on color synthesis of a plate roller image. A color multi-channel image is synthesized from a scanned plate roller image with halftone dots according to color settings and engraving processes to obtain a printing effect image of the plate roller. By comparing the image with the original, printing defects are reported. This method includes the following steps:

[0045] Step S1 : acquiring an image of the plate roller including halftone dots by performing a three-dimensional scanning on the plate roller, and establishing an actual halftone dot model.

[0046] Specifically, the plate roller to be inspected is first fixed to a dedicated plate roller fixture to ensure that the roller remains stable during the 3D scanning process. A high-precision 3D scanning device is then used to perform a full-scale scan of the fixed plate roller to obtain 3D data of the roller surface. The resolution of the 3D scanning device is configured to ensure that the tiny dot structure on the plate roller surface can be accurately captured. During the scanning process, the 3D scanning device rotates 360 degrees around the plate roller, collecting data every 0.5 degrees to obtain complete 3D information about the plate roller surface. The interval can be adjusted as needed.

[0047] The acquired 3D data is processed to generate a plate roller image containing dot information. This processing includes steps such as point cloud data registration, filtering, and meshing. Dots in the plate roller image are identified and extracted to create a real-world dot model. This model contains information such as the location, shape, size, and depth of each dot. To improve model accuracy, an edge detection algorithm is used to precisely locate dot boundaries, and morphological operations are used to optimize dot shape.

[0048] Step S2 : converting the pre-acquired original image into a theoretical dot model, and identifying dot defects based on the theoretical dot model and the actual dot model.

[0049] Specifically, this step includes the following sub-steps:

[0050] Step S201 : obtaining the position of each dot defect by comparing the difference between the theoretical dot model and the actual dot model.

[0051] First, a high-resolution digital image of the original print document is acquired, with a resolution of at least 300 dpi to ensure the integrity of image detail. Based on printing process parameters such as dot shape, dot size, and dot density, the original image is converted into a theoretical dot model. This conversion utilizes a digital halftoning method to convert the continuous-tone original image into a theoretical dot model composed of discrete dots. The theoretical dot model contains information such as the ideal position, shape, and size of each dot.

[0052] Step S202 : performing semantic segmentation on the original image, and obtaining object information where each dot defect is located based on the position of each dot defect.

[0053] Specifically, the location of each dot defect is determined by comparing the theoretical dot model with the actual dot model. This comparison utilizes image registration technology to spatially align the two models, then calculate the difference between the dots at corresponding locations. When the difference exceeds a preset threshold, a dot defect is identified. This threshold is determined based on print quality requirements and is typically set at 5% to 10% of the dot area.

[0054] Step S203 : sorting the dot defects based on the object information of each dot defect and the pre-built object-severity rules.

[0055] Semantic segmentation is performed on manuscript images, dividing them into distinct object regions, such as text, graphics, and background. Semantic segmentation uses deep learning models, such as U-Net or DeepLab, to classify images at the pixel level. These models, trained on extensive printed image data, are capable of accurately identifying different types of image objects.

[0056] Based on the location of each defect and the semantic segmentation results, the object information of each defect is obtained. Object information includes the type of object where the defect is located and its importance. For example, the defect may be located in the text area, graphic area, or background area.

[0057] Defects are ranked based on the object information for each defect and pre-established object-severity rules. Object-severity rules define the severity weights for defects in different object regions. For example, defects in text areas receive a weight of 0.9, defects in graphics areas receive a weight of 0.7, and defects in background areas receive a weight of 0.3. The final severity score for a defect is the product of its size and the corresponding object weight. Defects are sorted in descending order based on severity scores, with defects with higher scores receiving higher priority.

[0058] Step S3: predicting and obtaining a printing effect image after color synthesis based on the plate roller image.

[0059] Specifically, this step includes the following sub-steps:

[0060] Step S301: Calculate the volume of each dot based on the plate roller image.

[0061] The volume of each dot is calculated based on the plate roller image. The plate roller image is first preprocessed, including denoising, enhancement, and normalization. The dots are then segmented using a region growing or watershed algorithm to determine their boundaries. For each segmented dot, its three-dimensional volume is calculated. This volume calculation uses numerical integration to divide the dot region into tiny voxels, then accumulate the volumes of each voxel.

[0062] Step S302 : converting the volume of each dot into the saturation of each dot.

[0063] The volume of each dot is converted to the saturation of each dot. The conversion process uses a nonlinear mapping function, taking into account the physical properties of the ink and the printing process parameters. The mapping function is in the form of: S = a × (1-e -bV), where S represents saturation, V represents dot volume, and a and b are parameters determined based on ink characteristics. The values of parameters a and b vary for different ink colors and need to be determined through experimental data fitting.

[0064] Step S303 : obtaining a printing effect image by color synthesis based on the saturation of each dot and the pre-acquired ink-carrying color of each dot.

[0065] Based on the saturation of each dot and the previously acquired ink color, a printed image is generated through color synthesis. Color synthesis utilizes either an additive or subtractive color model, with the appropriate model selected based on the specific characteristics of the printing process. For four-color printing, the CMYK subtractive color model is used for color synthesis. The synthesis process takes into account ink overprinting and transparency, simulating the color performance of the actual printing process. The resulting printed image reflects the visual effect of the actual printing roller.

[0066] Step S4: Identify color defects based on the printed image and the original image. Specifically, the color defects include overprinting and moire.

[0067] Specifically, the printed image is registered with the original image to ensure spatial alignment. This registration process uses feature point matching and image transformation techniques to compensate for possible geometric differences such as rotation, scaling, and translation.

[0068] Calculates the difference between two images in color space. This difference is calculated using a color difference formula, such as CIEDE2000, that takes into account the human eye's perception of color differences. For each pixel, the color difference is calculated and a color difference map is generated.

[0069] Identify color defect areas based on the color difference chart and a preset threshold. When the color difference exceeds the threshold, it is considered a color defect. The threshold is set based on print quality requirements and is typically 3 to 5 color difference units.

[0070] Identified color defects are classified into overprint and moire defects. Overprint defects appear as misalignment between different color layers and are identified by analyzing the edge features in the color difference image. Moire defects appear as a network of uneven colors and are identified by analyzing the texture features in the color difference image.

[0071] Step S5: superimpose the color defects and the dot defects on the plate roller image, highlight the corresponding positions of the defects, and output them to the visualization terminal.

[0072] Specifically, the location information of identified dot defects and color defects is mapped onto the plate roller image. The mapping process takes into account the spatial correspondence between images to ensure the accuracy of defect locations.

[0073] Defective areas are highlighted to make them clearly visible on the plate roller image. Highlighting uses color enhancement technology to assign different highlight colors to different types of defects. For example, dot defects are marked in red, overprint defects are marked in blue, and moire defects are marked in green.

[0074] The processed images are output to a visualization terminal for operator review and analysis. The visualization terminal offers interactive functionality, supporting operations such as zooming in and out on defect areas and viewing detailed information. It also displays defect statistics, such as defect type, quantity, and severity ranking.

[0075] Example 2

[0076] This embodiment provides another method for detecting image defects based on color synthesis of a plate roller image, comprising the following steps:

[0077] Step S1, obtaining a plate roller image including halftone dots and establishing an actual halftone dot model.

[0078] Specifically, the plate roller to be inspected is first secured to a dedicated plate roller fixture to ensure it remains stable during the 3D scanning process. A high-precision 3D scanning device is then used to perform a full-scale scan of the secured plate roller, acquiring 3D data of its surface. The resolution of the 3D scanning device is configured to accurately capture the minute dot structure on the roller's surface. During the scanning process, the 3D scanning device rotates 360 degrees around the roller, collecting data every 0.3 degrees to obtain complete 3D information about the roller's surface.

[0079] The acquired 3D data is processed to generate a plate roller image containing dot information. This processing includes steps such as point cloud data registration, filtering, and meshing. Dots in the plate roller image are identified and extracted to create a real-world dot model. This model contains information such as the location, shape, size, and depth of each dot. To improve model accuracy, an improved Canny edge detection algorithm is used to precisely locate dot boundaries, and morphological operations are used to optimize dot shape.

[0080] Step S2 : converting the pre-acquired original image into a theoretical dot model, and identifying dot defects based on the theoretical dot model and the actual dot model.

[0081] Specifically, a high-resolution digital image of the printed manuscript is first acquired, with a resolution of at least 400 dpi to ensure the integrity of image details. Based on printing process parameters such as dot shape, dot size, and dot density, the manuscript image is converted into a theoretical dot model. This conversion process utilizes an improved digital halftoning algorithm to convert the continuous-tone manuscript image into a theoretical dot model composed of discrete dots. The theoretical dot model contains information such as the ideal position, shape, and size of each dot.

[0082] By comparing the theoretical dot model with the actual dot model, the location of each dot defect is determined. This comparison utilizes feature point matching and image registration techniques to spatially align the two models. The difference between the dots at corresponding locations is then calculated. When the difference exceeds a preset threshold, a dot defect is identified. This threshold is determined based on print quality requirements and is typically set between 3% and 8% of the dot area.

[0083] Semantic segmentation is performed on manuscript images, dividing them into distinct object regions, such as text, graphics, and background. Semantic segmentation uses improved deep learning models, such as Mask R-CNN or PSPNet, to classify images at the pixel level. These models, trained on a large amount of printed image data, are capable of accurately identifying different types of image objects.

[0084] Based on the location of each defect and the semantic segmentation results, the object information of each defect is obtained. Object information includes the type of object where the defect is located and its importance. For example, the defect may be located in the text area, graphic area, or background area.

[0085] Defects are ranked based on the object information for each defect and pre-established object-severity rules. Object-severity rules define the severity weights for defects in different object regions. For example, defects in text areas receive a weight of 0.95, defects in graphics areas receive a weight of 0.75, and defects in background areas receive a weight of 0.25. The final severity score for a defect is the product of its size and the corresponding object weight. Defects are sorted in descending order based on severity scores, with higher-scoring defects receiving higher priority.

[0086] Step S3: predicting a printing effect image after color synthesis based on the plate roller image.

[0087] Specifically, the volume of each dot is calculated based on the plate cylinder image. First, preprocess the plate cylinder image, including operations such as denoising, enhancement, and normalization. Then use an improved watershed algorithm to segment the dots and obtain the boundary of each dot. For each segmented dot, calculate its three-dimensional volume. The volume calculation uses an adaptive numerical integration method, divides the dot area into voxels of different sizes, dynamically adjusts the voxel size according to the regional complexity, and then accumulates the volumes of each voxel.

[0088] Convert the volume of each dot to the saturation of each dot. The conversion process uses a piecewise non-linear mapping function, considering the physical properties of the ink and the printing process parameters. The form of the mapping function is: when V < V0, S = c × V; when V ≥ V0, where S represents saturation, V represents dot volume, and V0, a, b, and c are parameters determined according to the ink properties. For inks of different colors, the values of these parameters are different and need to be obtained by fitting experimental data.

[0089] Based on the saturation of each dot and the ink-carrying color of each dot obtained in advance, obtain the printed effect image through color synthesis. Color synthesis uses an improved CMYK subtractive model, considering the interaction between inks and paper properties. The synthesis process simulates the color overprint effect in the actual printing process, including physical processes such as the transparency, diffusion, and absorption of the ink. The finally generated printed effect image is closer to the visual effect after actual printing.

[0090] Step S4, identify color defects based on the printed effect image and the original manuscript image.

[0091] Specifically, register the printed effect image with the original manuscript image to ensure that the two images are spatially aligned. The registration process uses multi-scale feature point matching and non-rigid image transformation techniques to compensate for possible geometric deformations.

[0092] Calculate the difference between the two images in the color space. The difference calculation uses an improved color difference formula, such as CIEDE2000 or DeltaE94, considering the human eye's perception characteristics of different color differences. For each pixel position, calculate its color difference value and generate a color difference map.

[0093] According to the color difference map and the adaptive threshold, identify the color defect area. The adaptive threshold is dynamically adjusted according to the characteristics of the image area. A lower threshold is used for important areas, and a higher threshold is used for secondary areas. The reference value of the threshold is determined according to the printing quality requirements and is usually 2 to 6 color difference units.

[0094] Identified color defects are classified into overprint and moire defects. Overprint defects appear as misalignment between different color layers and are identified by analyzing the edge features and directionality in the color difference image. Moire defects appear as a network-like texture with uneven color and are identified by analyzing the texture features and frequency characteristics in the color difference image.

[0095] Step S5: superimposing the color defect and the halftone dot defect on the plate roller image, highlighting the corresponding position of the defect, and outputting it to a visualization terminal.

[0096] Specifically, the location information of identified dot defects and color defects is mapped onto the plate roller image. The mapping process takes into account the spatial correspondence and deformation characteristics between images to ensure the accuracy of defect locations.

[0097] Highlight defective areas to make them clearly visible on the cylinder image. Highlighting uses adaptive color enhancement technology to dynamically adjust the highlight color based on the background color, ensuring that defects appear clearly against any background. Different highlight colors and styles are assigned to different defect types and severities. For example, severe dot defects are marked with a solid red line, minor dot defects with a dashed orange line, overprint defects with a blue line, and moire defects with a green line.

[0098] The processed images are output to a high-resolution visualization terminal for operator review and analysis. The visualization terminal features enhanced interactive functionality, supporting multi-level magnification of defect areas, comparative analysis, and detailed information viewing. Furthermore, defect statistics and distribution maps, such as defect type distribution and severity heat maps, are displayed on the visualization terminal, helping operators quickly locate and address critical defects.

[0099] Example 3

[0100] Based on Example 1, this embodiment provides a graphic defect detection system for color synthesis based on a plate roller image, including:

[0101] Plate roller fixing module, used to fix the plate roller to be tested;

[0102] A three-dimensional scanning module is used to perform three-dimensional scanning on the fixed plate roller to obtain an image of the plate roller including dots;

[0103] A data processing module, configured to detect defects based on the plate roller image by using the image-text defect detection method for color synthesis based on the plate roller image as described in Example 1;

[0104] a database storing object-severity rules for ranking dot defects;

[0105] Visualization terminal, used to display the results of defect identification.

[0106] Specifically, the plate roller fixing module includes an adjustable bracket and clamp to accommodate rollers of varying sizes and weights. The bracket is constructed from a high-strength alloy, ensuring excellent stability and vibration resistance. The clamp is pneumatically controlled, ensuring uniform force and avoiding damage to the roller surface. The fixing module also features a level and fine-tuning mechanism to ensure the roller remains level and centered during the fixing process.

[0107] The 3D scanning module utilizes high-precision structured light scanning technology and consists of a projection unit and multiple image acquisition units. The projection unit projects a structured grating pattern, while the image acquisition units capture reflected light from various angles. Mounted on a precision motion platform, the scanning module can move axially and radially along the plate roller, enabling full-surface scanning. Scanning resolution reaches 10 microns, enabling the capture of detailed dot details. During the scanning process, the system automatically controls ambient light and temperature to minimize the impact of external factors on scan quality.

[0108] The data processing module includes a high-performance computing server and dedicated image processing software. The computing server is equipped with a multi-core CPU and GPU accelerator card, enabling efficient processing of large amounts of 3D and image data. The image processing software integrates functional modules such as point cloud processing, image segmentation, feature extraction, and defect recognition. The data processing module implements the defect detection method described in Example 1, including steps such as establishing an actual dot model, converting a theoretical dot model, identifying dot defects, color synthesis, and identifying color defects.

[0109] The database utilizes a distributed storage architecture, comprising both relational and non-relational databases. The relational database stores structured data, such as object-severity rules, plate roller parameters, and inspection results. The non-relational database stores large amounts of image data and 3D model data. Object-severity rules are customized for different printing applications, including weights and thresholds for different object types. The database supports dynamic rule updates and version management, facilitating continuous system optimization and adaptation to diverse inspection needs.

[0110] The visualization terminal includes a high-resolution display and interactive control unit. The display utilizes professional-grade color calibration technology to ensure accurate color display. The interactive control unit includes a touch screen, mouse, and keyboard, supporting multiple operation modes. The visualization terminal provides an intuitive user interface that displays defect detection results, including information such as defect location, type, and severity. Users can use the interface to view detailed information at different levels, such as the original plate image, dot model, and color synthesis results. The terminal also provides a report generation function, allowing inspection results to be exported for archiving and analysis.

[0111] Example 3

[0112] This embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the graphic defect detection method for color synthesis based on the plate roller image as described in Example 1.

[0113] Specifically, the electronic device adopts a modular design, including a main control unit, an image acquisition unit, a data processing unit, and a display interaction unit. The main control unit is responsible for coordinating the work of each unit to ensure the stability and reliability of the system operation.

[0114] The processor consists of a central processing unit (CPU) and a graphics processing unit (GPU). The CPU utilizes a multi-core architecture with a clock speed of at least 3.5GHz, supporting efficient parallel computing. The GPU is dedicated to image processing and deep learning computations, equipped with at least 8GB of video memory and supporting CUDA or OpenCL acceleration. The processors are connected via a high-speed bus, forming a heterogeneous computing platform capable of efficiently processing large-scale image data and complex algorithms.

[0115] Memory includes RAM (RAM) and storage devices. RAM capacity should be at least 32GB, using DDR4 or higher specifications, to ensure smooth system operation. Storage devices include solid-state drives (SSDs) and hard disk drives (HDDs). The SSD is used for fast loading of the operating system and commonly used programs, with a capacity of at least 512GB; the HDD is used for large-capacity data storage, with a capacity of at least 2TB. The storage system uses RAID technology to improve data security and read and write speeds.

[0116] The program includes an operating system, basic libraries, and application software. The operating system uses a Linux or Windows system with good real-time performance and is optimized for image processing tasks. The basic libraries include image processing libraries (such as OpenCV), deep learning frameworks (such as TensorFlow or PyTorch), and numerical computing libraries (such as NumPy). The application software implements the defect detection method described in Example 1 and includes modules such as image preprocessing, dot model establishment, defect identification, and result visualization.

[0117] The electronic devices are also equipped with input and output interfaces, including USB, HDMI, and network interfaces, to facilitate connection with external devices and data exchange. The modular design of the equipment facilitates upgrades and maintenance. The power supply system is equipped with an uninterruptible power supply (UPS) to prevent data loss caused by unexpected power outages. The cooling system uses a combination of air cooling and heat pipes to ensure that the equipment maintains a stable temperature during long-term operation.

[0118] Example 4

[0119] This embodiment provides a computer-readable storage medium, including one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the graphic defect detection method for color synthesis based on the plate roller image as described in Example 1.

[0120] Specifically, computer-readable storage media can include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. The storage media utilizes a layered storage architecture, including a system layer, an algorithm layer, and an application layer. The system layer stores the operating system and driver programs, the algorithm layer stores the basic algorithm library and model files, and the application layer stores the program code and configuration files that implement the defect detection method.

[0121] The program code in the storage medium adopts a modular design, including image acquisition module, pre-processing module, dot model building module, defect recognition module, and result visualization module. Data is exchanged between modules via standard interfaces, ensuring the scalability and maintainability of the system.

[0122] The image acquisition module is responsible for controlling the 3D scanning device to acquire 3D data from the plate roller. The module includes device drivers and parameter configuration functions, supporting different types of scanning devices. During the acquisition process, the module monitors data quality in real time and automatically adjusts scanning parameters to ensure high-quality plate roller images.

[0123] The preprocessing module processes the collected 3D data, including operations such as point cloud registration, filtering, and gridding. The module utilizes multi-threaded processing technology to improve data processing efficiency. The processed data is saved in a standard format for subsequent analysis and processing.

[0124] The dot modeling module extracts dot information from the processed data and builds a practical dot model. Simultaneously, the module converts the original image into a theoretical dot model. Both models utilize a unified data structure, facilitating subsequent comparison and analysis.

[0125] The defect recognition module identifies dot and color defects. It integrates algorithms such as image registration, difference calculation, and threshold segmentation. For dot defects, the module also implements defect sorting based on object information. For color defects, the module classifies defects such as overprint and moire.

[0126] The results visualization module presents identification results in an intuitive manner, including defect location marking and statistical information display. The module supports various interactive operations, such as defect area magnification and comparative analysis. It also provides a report generation function that can export inspection result reports.

[0127] The storage medium also contains configuration files and model files. Configuration files store system and algorithm parameters, facilitating adjustments based on different application scenarios. Model files store pre-trained deep learning models, such as neural network models for semantic segmentation. These files are stored in a standard format for easy updating and management.

[0128] It should be noted that Example 1, Example 2, Example 3, Example 4, and Example 5 are all a method for detecting graphic defects based on color synthesis of a plate roller image.

[0129] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A method for detecting image defects based on color synthesis of plate roller images, characterized in that: The steps include: Step S1, obtaining an image of a plate roller including halftone dots and establishing an actual halftone dot model, wherein the image of the plate roller is obtained by performing a three-dimensional scanning on the plate roller; Step S2, converting the pre-acquired original image into a theoretical dot model, and identifying dot defects based on the theoretical dot model and the actual dot model; Step S3, predicting a printing effect image after color synthesis based on the plate roller image; Step S4: identifying color defects based on the printing effect image and the original image.

2. The method for detecting image defects based on color synthesis of plate roller images according to claim 1, characterized in that: The step S3 includes the following sub-steps: Step S301, calculating the volume of each dot based on the plate roller image; Step S302, converting the volume of each dot into the saturation of each dot; Step S303 : obtaining a printing effect image by color synthesis based on the saturation of each dot and the pre-acquired ink-carrying color of each dot.

3. The method for detecting image defects based on color synthesis of plate roller images according to claim 1, characterized in that: The step S2 includes the following sub-steps: Step S201, obtaining the position of each dot defect by comparing the difference between the theoretical dot model and the actual dot model; Step S202 , performing semantic segmentation on the original image, and obtaining object information where each dot defect is located based on the location of each dot defect; Step S203 : sorting the dot defects based on the object information of each dot defect and the pre-built object-severity rules.

4. The method for detecting image defects based on color synthesis of plate roller images according to claim 1, characterized in that: Also includes: Step S5: superimpose the color defect and the halftone dot defect on the plate roller image, highlight the position corresponding to the defect, and output it to a visualization terminal.

5. The method for detecting image defects based on color synthesis of plate roller images according to claim 1, characterized in that: The color defects mentioned include overprint and moire.

6. A graphic defect detection system based on color synthesis of plate roller images, characterized in that: include: Plate roller fixing module, used to fix the plate roller to be tested; A three-dimensional scanning module is used to perform three-dimensional scanning on the fixed plate roller to obtain an image of the plate roller including dots; A data processing module is used to realize defect detection based on the plate roller image by the image and text defect detection method based on color synthesis of the plate roller image according to any one of claims 1 to 5.

7. The image defect detection system based on color synthesis of plate roller images according to claim 6, characterized in that: Also includes: The database stores object-severity rules for ranking dot defects.

8. The image defect detection system based on color synthesis of plate roller images according to claim 6, characterized in that: Also includes: Visualization terminal, used to display the results of defect identification.

9. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory stores one or more programs, wherein the one or more programs include instructions for executing the method for detecting image defects by color synthesis based on a plate roller image as described in any one of claims 1-5.

10. A computer-readable storage medium, characterized in that It includes one or more programs for execution by one or more processors of an electronic device, and the one or more programs include instructions for executing the graphic defect detection method for color synthesis based on the plate roller image as described in any one of claims 1-5.

Images