Image Recognition-Based Method for Detecting Printing Accuracy of Color and Pattern Designs on Color-Coated Steel Sheets

Abstract

This invention discloses a method for detecting the printing accuracy of patterns on color-coated steel sheets based on image recognition, belonging to the field of printing accuracy detection technology. The method includes acquiring multi-dimensional influence data on printing production, and determining whether the printing environment for the color-coated steel sheet's patterns is normal based on this data. If the printing environment is abnormal, printing is stopped. If the environment is normal, the printed color-coated steel sheet is acquired, its image is extracted and preprocessed to obtain a preprocessed image. Pattern processing is performed on the image to obtain the color-coated steel sheet's patterns, and a printing accuracy error coefficient is obtained. A printing accuracy error threshold is determined based on the multi-dimensional influence data, and the printing accuracy is evaluated using the printing accuracy error coefficient. This method solves the problem that existing methods for detecting the printing accuracy of color-coated steel sheet patterns are difficult to comprehensively and accurately detect.

Description

Technical Field

[0001] This invention relates to the field of printing accuracy testing technology, specifically to a method for testing the printing accuracy of color patterns on color-coated steel plates based on image recognition. Background Technology

[0002] There are various methods for inspecting the printing accuracy of patterns on color-coated steel sheets, each with its own characteristics and suitable for different scenarios and accuracy requirements. Common methods include visual inspection, dimensional measurement, optical inspection, and image analysis. Among them, image analysis methods include those based on scanning imaging and software analysis, as well as those based on artificial intelligence image recognition and detection.

[0003] The scanning imaging and software analysis method involves scanning or photographing the printed pattern on the color-coated steel sheet using a scanner or high-resolution camera to acquire a digital image. Then, specialized image analysis software is used to process and analyze the image. The software can automatically identify pattern features, measure various parameters such as line position accuracy, pattern repeatability, and color uniformity, and compare them with pre-set standard images or design files to quickly and accurately calculate deviation values ​​and generate a detailed inspection report.

[0004] The AI-based image recognition and detection method utilizes deep learning algorithms to train a large number of images of color-coated steel plate printing patterns, allowing the computer model to learn the characteristics of normal and defective patterns. In actual inspection, the model can automatically identify various defects and deviations in the patterns, and can even classify and statistically analyze different types of defects.

[0005] Existing image analysis methods for printing accuracy detection have limitations in comprehensively and accurately detecting the printing accuracy of patterns and designs on color-coated steel sheets by considering factors such as printing equipment, workshop environment, coating characteristics, and the characteristics of the color-coated steel sheet itself. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method for detecting the printing accuracy of patterns on color-coated steel sheets based on image recognition, which solves the problem that existing methods for detecting the printing accuracy of patterns on color-coated steel sheets are difficult to comprehensively and accurately detect.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting the printing accuracy of color-coated steel plate patterns based on image recognition, comprising the following steps: acquiring multi-dimensional influence data on printing production, and determining whether the printing environment for color-coated steel plate patterns is normal based on the multi-dimensional influence data; if the printing environment for color-coated steel plate patterns is abnormal, printing is stopped; if the printing environment for color-coated steel plate patterns is normal, the printed color-coated steel plate is acquired, the color-coated steel plate image is extracted and preprocessed to obtain the preprocessed color-coated steel plate image; pattern processing is performed on the color-coated steel plate image to obtain the color-coated steel plate pattern, and the printing accuracy error coefficient is obtained; the printing accuracy error threshold is determined based on the multi-dimensional influence data on printing production, and the printing accuracy is evaluated in conjunction with the printing accuracy error coefficient.

[0008] Furthermore, the multidimensional impact data on printing production includes printing equipment operation data, printing workshop environmental parameter data, coating characteristic parameter data, and color-coated plate characteristic parameter data.

[0009] Furthermore, determining whether the printing environment for color-coated steel plate patterns is normal based on multi-dimensional influence data generated during printing includes the following steps: Using printing equipment operation data as input to a support vector machine model to obtain the printing equipment's operating status; if the printing equipment's operating status is abnormal, the printing environment for color-coated steel plate patterns is abnormal. If the printing equipment's operating status is normal, then using printing workshop environmental parameter data as input to an environmental factor judgment model built based on a convolutional neural network to obtain the printing workshop's environmental status; if the printing workshop's environmental status is abnormal, the printing environment for color-coated steel plate patterns is abnormal. If the printing workshop's environmental status is normal, then using coating characteristic parameter data and color-coated steel plate characteristic parameter data as input to a random forest model to obtain the coating characteristic and color-coated steel plate characteristic status; if the coating characteristic and color-coated steel plate characteristic status are abnormal, the printing environment for color-coated steel plate patterns is abnormal; if the coating characteristic and color-coated steel plate characteristic status are normal, the printing environment for color-coated steel plate patterns is normal.

[0010] Further, the color-coated steel plate image is processed to obtain a color-coated steel plate pattern, including the following steps: edge detection is performed on the color-coated steel plate image to obtain an edge contour image; corner detection is performed on the edge contour image based on a corner detection algorithm to obtain a set of corner coordinates; threshold segmentation is performed based on the edge contour image and the set of corner coordinates to obtain the color-coated steel plate pattern.

[0011] Further, edge detection is performed on the color-coated steel plate image to obtain an edge contour image, including the following steps: After smoothing the color-coated steel plate image, a gradient image is obtained based on the Sobel operator; non-maximum suppression is applied to the gradient image to obtain a thinned edge image; the initial double-threshold detection interval stored in the database is obtained; the printing workshop environment-double-threshold correction coefficient mapping database stored in the database is obtained, and the corresponding double-threshold correction coefficient is retrieved based on the current printing workshop environment parameter data; the initial double-threshold detection interval is corrected based on the double-threshold correction coefficient to obtain a corrected double-threshold detection interval; the pixel value of each pixel in the thinned edge image is obtained: if the pixel value of a certain pixel is greater than the maximum value in the corrected double-threshold detection interval, the pixel is identified as an edge pixel; if the pixel value of a certain pixel is less than the minimum value in the corrected double-threshold detection interval, the pixel is identified as a non-edge pixel; if the pixel value of a certain pixel is within the corrected double-threshold detection interval, it is determined whether the pixel is connected to an edge pixel; if connected, it is identified as an edge pixel; if not connected, it is identified as a non-edge pixel; all edge pixels are connected to obtain an edge contour image.

[0012] Further, threshold segmentation is performed based on the edge contour image and the set of corner coordinates to obtain the color pattern of the color-coated steel plate. This includes the following steps: obtaining the pixel grayscale values ​​of all pixels in the edge contour image, drawing the grayscale value distribution, determining the most concentrated grayscale interval, and defining the most concentrated grayscale interval as the initial threshold segmentation range; traversing the pixel grayscale values ​​of the neighborhood pixels of each corner point in the set of corner coordinates, obtaining the mean and standard deviation; if the mean is lower than the mean of the pixel grayscale values ​​of all pixels in the edge contour image and the standard deviation is greater than the set standard deviation threshold, then contaminant data is obtained, and the initial threshold segmentation range is corrected based on the contaminant data to obtain the corrected threshold segmentation range; otherwise, the initial threshold segmentation range is not corrected; if the pixel grayscale value of a certain pixel in the color-coated steel plate image is within the corrected threshold segmentation range or the uncorrected initial threshold segmentation range, then the pixel is recorded as the pattern area; otherwise, it is recorded as the background area, thus obtaining the color pattern of the color-coated steel plate.

[0013] Further, obtaining the printing accuracy error coefficient includes the following steps: obtaining the standard pattern of the color-coated steel plate stored in the database, and aligning the color-coated steel plate pattern with the standard pattern; obtaining the dimensional error Cw, positional error Ww, and color error Yw; and obtaining the printing accuracy error coefficient Yx based on the error accuracy evaluation formula.

[0014] Yx = ω1*Cw + ω2*Ww + ω3*Yw;

[0015] Where ω1 is the weighting factor of Cw, ω2 is the weighting factor of Ww, and ω3 is the weighting factor of Yw.

[0016] Furthermore, the printing accuracy error threshold is determined based on multi-dimensional impact data of printing production, including the following steps: Printing equipment operation data and printing workshop environmental parameter data are used as inputs to a decision tree model to obtain the theoretical size error Clw and the theoretical position error Wlw; coating characteristic parameter data and printing workshop environmental parameter data are used as inputs to a multiple linear regression model to obtain the first color theoretical error fYlw; color-coated sheet characteristic parameter data and printing workshop environmental parameter data are used as inputs to a random forest regression model to obtain the second color theoretical error sYlw; based on the theoretical size error Clw, the theoretical position error Wlw, the first color theoretical error fYlw, and the second color theoretical error sYlw, the printing accuracy error threshold Yxy is obtained.

[0017]

[0018] Where ω1′ is the weighting factor of Clw, ω2′ is the weighting factor of Wlw, and ω3′ is... Weighting factors.

[0019] Furthermore, the printing accuracy is evaluated by combining the printing accuracy error coefficient, including the following steps: if the printing accuracy error coefficient is less than the printing accuracy error threshold, the accuracy test result is normal; if the printing accuracy error coefficient is not less than the printing accuracy error threshold, the accuracy test result is abnormal, and the multi-dimensional impact data of printing production is optimized and controlled.

[0020] Furthermore, the optimization and control of multi-dimensional impact data on printing production includes the following steps: establishing a database mapping dimensional position error - equipment and environment data based on the input and output of a decision tree model; establishing a database mapping first color error - coating and environment data based on the input and output of a random forest regression model; establishing a database mapping second color error - color-coated plate and environment data based on the input and output of a random forest regression model; and obtaining the error allowable thresholds stored in the databases, including the allowable threshold for dimensional error, the allowable threshold for position error, the allowable threshold for first color error, and the allowable threshold for second color error.

[0021] The data sets of multidimensional impact data on printing production to be evaluated are selected from the dimensional position error-equipment and environment data mapping database, the first color error-coating and environment data mapping database, and the second color error-color coated plate and environment data mapping database.

[0022] Based on the combination of multidimensional impact data on printing production for each candidate to be evaluated, the corresponding comprehensive error value dYx for each candidate to be evaluated is obtained; the cost-benefit ratio Xyb and the feasibility score Pf for each combination of multidimensional impact data on printing production for each candidate to be evaluated are determined.

[0023] Determine the various evaluation coefficients (px);

[0024]

[0025] Where β1 is The weighting factors are β2 and β3 respectively. The weighting factor, e is the natural constant;

[0026] The combination of multidimensional impact data on printing production corresponding to the highest evaluation coefficient is used as the adjustment target to optimize and control the multidimensional impact data on printing production.

[0027] The present invention has the following beneficial effects:

[0028] This image recognition-based method for detecting the printing accuracy of patterns on color-coated steel sheets comprehensively considers the impact of various factors such as equipment, environment, coatings, and color-coated steel sheets on printing accuracy. It can promptly detect abnormalities in the printing process, stop printing when abnormal printing environments occur to avoid producing defective products, and quantitatively judge the printing accuracy based on the evaluation results, providing a basis for subsequent optimization and control. This improves the printing quality of color-coated steel sheets, reduces production costs, and solves the problem of the difficulty in comprehensively and accurately detecting the printing accuracy of patterns on color-coated steel sheets in existing technologies.

[0029] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0030] Figure 1 This is a flowchart of the image recognition-based method for detecting the printing accuracy of color patterns on color-coated steel sheets according to the present invention.

[0031] Figure 2 This is a flowchart illustrating the process of evaluating printing accuracy using the image recognition-based color-coated steel plate printing accuracy detection method of the present invention. Detailed Implementation

[0032] Please see Figure 1 This invention provides a technical solution: a method for detecting the printing accuracy of color patterns on color-coated steel sheets based on image recognition, comprising the following steps: acquiring multi-dimensional influence data on printing production, and determining whether the printing environment for color patterns on color-coated steel sheets is normal based on the multi-dimensional influence data generated during printing.

[0033] Multidimensional data affecting printing production include printing equipment operation data (including printing roller pressure, printing speed, roller diameter deviation, transmission system accuracy, and positioning device error), printing workshop environmental parameter data (temperature and humidity, light intensity, air quality, vibration, and airflow), coating characteristic parameter data (pigment particle size, pigment concentration, and drying speed), and color-coated plate characteristic parameter data (surface roughness, material's adsorption of coating, and coating thickness).

[0034] Using printing equipment operating data as input to a support vector machine (SVM) model, the operating status of the printing equipment is obtained. Stable operation of the printing equipment is fundamental to ensuring printing accuracy. By analyzing equipment operating data such as printing roller pressure and printing speed through the SVM model, it is possible to quickly and accurately determine whether the equipment is in normal working condition.

[0035] If the printing equipment is operating abnormally, the printing environment for the color and pattern of the color-coated steel sheet is also abnormal. If the printing equipment is operating normally, the environmental parameter data of the printing workshop is used as input to an environmental factor judgment model built based on a convolutional neural network to obtain the environmental status of the printing workshop. The workshop environment has a significant impact on printing quality. Changes in environmental parameters such as temperature, humidity, and light intensity will alter the drying speed of the coating and the physical properties of the color-coated steel sheet. Analyzing these parameters using a model built using a convolutional neural network can promptly detect environmental anomalies.

[0036] If the printing workshop environment is abnormal, the printing environment for the color-coated steel plate patterns will also be abnormal. If the printing workshop environment is normal, the paint characteristic parameter data and the color-coated steel plate characteristic parameter data will be used as input to the random forest model to obtain the paint and color-coated steel plate characteristic states. The characteristics of the paint and the color-coated steel plate themselves are key factors determining printing accuracy. Parameters such as pigment particle size, concentration, surface roughness of the color-coated steel plate, and its adsorption of paint will all affect the pattern presentation. Analyzing these characteristic parameters using the random forest model can effectively determine whether they are normal.

[0037] If the characteristics of the coating and the color-coated sheet are abnormal, then the printing environment for the color-coated sheet is abnormal; if the characteristics of the coating and the color-coated sheet are normal, then the printing environment for the color-coated sheet is normal.

[0038] If the printing environment for the color-coated steel sheet pattern is abnormal, printing is stopped; if the printing environment is normal, the printed color-coated steel sheet is acquired, its image is extracted and preprocessed, and a preprocessed image is obtained. Actual acquired color-coated steel sheet images may have noise interference and uneven lighting, which can affect subsequent pattern analysis and feature extraction. Preprocessing operations (such as denoising, grayscale conversion, and histogram equalization) can effectively improve image quality, remove noise, enhance image contrast and clarity, and make pattern features in the image more obvious. This provides a high-quality base image for subsequent edge detection, corner detection, and other operations, thereby improving the accuracy and reliability of the entire detection process.

[0039] Edge detection is performed on the color-coated steel plate image to obtain the edge contour image; edges are important features of the pattern in the image, containing the shape and structural information of the pattern. Edge detection can accurately locate the boundary of the pattern and separate the pattern from the background.

[0040] After smoothing the color-coated plate image, a gradient image is obtained based on the Sobel operator. The Sobel operator can calculate the gradient of the image in the horizontal and vertical directions, highlighting areas with obvious gray-scale changes, i.e. potential pattern edges, and providing preliminary gradient information for subsequent accurate edge identification.

[0041] Non-maximum suppression (NMS) is applied to the gradient image to obtain a thinned edge image. NMS only retains the local maximum pixels in the gradient direction and removes other redundant pixels, thus thinning the edges and making them more accurately reflect the pattern boundaries. This reduces the error in edge position judgment in subsequent processing and improves the accuracy of edge detection.

[0042] The process involves retrieving the initial double-threshold detection interval from the database; retrieving the printing workshop environment-double-threshold correction coefficient mapping database from the database; and retrieving the corresponding double-threshold correction coefficients based on the current printing workshop environment parameters; then correcting the initial double-threshold detection interval based on the double-threshold correction coefficients (multiplying the coefficients by the interval endpoint values) to obtain the corrected double-threshold detection interval. Different printing workshop environments (temperature, humidity, lighting, etc.) affect image quality and edge features. Retrieving the initial double-threshold detection interval from the database is the basic basis for edge judgment. Combining the current printing workshop environment parameters with retrieving the coefficient correction threshold interval from the "printing workshop environment-double-threshold correction coefficient mapping database" can adapt to different environmental changes and make edge detection more accurate.

[0043] Obtain the pixel value of each pixel in the refined edge image: If the pixel value of a certain pixel is greater than the maximum value in the corrected double threshold detection interval, the pixel is identified as an edge pixel; if the pixel value of a certain pixel is less than the minimum value in the corrected double threshold detection interval, the pixel is identified as a non-edge pixel; if the pixel value of a certain pixel is within the corrected double threshold detection interval, determine whether the pixel is connected to an edge pixel. If connected, it is identified as an edge pixel; if not connected, it is identified as a non-edge pixel. Connect all edge pixels to obtain the edge contour image.

[0044] Corner detection algorithms (Harris corner detection algorithm or SIFT feature point detection algorithm) are used to detect corners in edge contour images, obtaining a set of corner coordinates. Corners are points with distinctive features in an image, typically located at the corners of a pattern or the intersections of edges. Corners possess unique geometric features and stability, providing important reference points for pattern matching and localization. By detecting the set of corner coordinates, key positions of the pattern can be determined, further improving the accuracy of pattern position detection.

[0045] Thresholding segmentation, based on edge contour images and corner coordinate sets, yields the color patterns of precast steel plates. Thresholding segmentation is a simple and effective image segmentation method that divides an image into pattern regions and background regions by comparing the grayscale or color values ​​of pixels with a set threshold. Combining edge contour images and corner coordinate sets for thresholding segmentation allows for more accurate determination of the pattern's extent, removal of background noise interference, and the acquisition of clear precast steel plate color patterns.

[0046] The process involves acquiring the grayscale values ​​of all pixels within the edge contour image, plotting the grayscale value distribution, identifying the most concentrated grayscale interval, and defining this interval as the initial threshold segmentation range. By acquiring the grayscale values ​​of all pixels and plotting their distribution, the concentrated trend of image grayscale can be visually presented. Setting the most concentrated grayscale interval as the initial threshold range effectively distinguishes patterns from the background. Because the grayscale values ​​of pattern areas typically exhibit similarity and concentration, this method can initially delineate the pattern's range.

[0047] The algorithm iterates through the pixel grayscale values ​​of the neighborhood pixels of each corner point in the corner coordinate set, obtaining the mean and standard deviation. If the mean is lower than the mean of pixel grayscale values ​​of all pixels in the edge contour image and the standard deviation is greater than the set standard deviation threshold, then contaminant data is obtained. The initial threshold segmentation range is then corrected based on the contaminant data to obtain the corrected threshold segmentation range. Otherwise, the initial threshold segmentation range is not corrected (by retrieving the contaminant-threshold segmentation range correction coefficient mapping database stored in the database, retrieving the corresponding threshold segmentation range correction coefficient based on the current contaminant data, and correcting the initial threshold segmentation range based on the threshold segmentation range correction coefficient, multiplying the coefficient by the interval endpoint value). Based on this, contaminant data is obtained and the threshold range is corrected, which can adapt to complex image situations and more accurately define the pattern area.

[0048] If the grayscale value of a pixel in the color-coated plate image is within the corrected threshold segmentation range or the uncorrected initial threshold segmentation range, then the pixel is recorded as the pattern area; otherwise, it is recorded as the background area, thus obtaining the color pattern of the color-coated plate.

[0049] The image of the color-coated steel plate is processed to obtain the color pattern of the color-coated steel plate, and the printing accuracy error coefficient is obtained.

[0050] Obtaining the printing accuracy error coefficient includes the following steps: obtaining the standard pattern of color-coated steel plate stored in the database, and aligning the color-coated steel plate pattern with the standard pattern; obtaining the dimensional error Cw, positional error Ww, and color error Yw; and obtaining the printing accuracy error coefficient Yx based on the error accuracy evaluation formula.

[0051] Yx = ω1*Cw + ω2*Ww + ω3*Yw;

[0052] Where ω1 is the weighting factor of Cw, ω2 is the weighting factor of Ww, and ω3 is the weighting factor of Yw.

[0053] Size error reflects the difference between the pattern and the standard in terms of size; positional error reflects the offset of the pattern in spatial position; and color error shows the degree of deviation of the pattern's color from the standard. Comprehensive acquisition of these three error data allows for the measurement of print quality from multiple dimensions.

[0054] The printing accuracy error threshold is determined based on multidimensional impact data of printing production, and the printing accuracy is evaluated by combining the printing accuracy error coefficient.

[0055] Using printing equipment operating data and printing workshop environmental parameters as inputs to the decision tree model, the theoretical error of size (Clw) and the theoretical error of position (Wlw) are obtained. Printing equipment operating data (such as printing roller pressure and speed) and workshop environmental parameters (temperature, humidity, vibration, etc.) have a significant impact on the accuracy of pattern size and position. Based on historical data and equipment and environmental parameters, the decision tree model predicts the theoretical error range of size and position, and can uncover the relationships between these complex data points.

[0056] Using coating characteristic parameters and printing workshop environmental parameters as inputs to a multiple linear regression model, the first theoretical color error fYlw is obtained. Coating characteristics (pigment particle size, concentration, etc.) and environmental parameters jointly influence the color of the printed pattern. The multiple linear regression model can quantitatively analyze their relationship with color error, calculating the first theoretical color error based on current coating and environmental data.

[0057] Using the characteristic parameters of the pre-coated steel sheet and the environmental parameters of the printing workshop as inputs to the random forest regression model, the theoretical error sYlw for the second color is obtained. The characteristics of the pre-coated steel sheet (surface roughness, adsorption, etc.) also affect the color effect. The random forest regression model can comprehensively consider the characteristics of the pre-coated steel sheet and environmental parameters to further analyze the color error.

[0058] Based on the theoretical size error Clw, the theoretical position error Wlw, the theoretical first color error fYlw, and the theoretical second color error sYlw, the printing accuracy error threshold Yxy is obtained:

[0059]

[0060] Where ω1′ is the weighting factor of Clw, ω2′ is the weighting factor of Wlw, and ω3′ is... Weighting factors.

[0061] like Figure 2As shown, if the printing accuracy error coefficient is less than the printing accuracy error threshold, the accuracy test result is normal; if the printing accuracy error coefficient is not less than the printing accuracy error threshold, the accuracy test result is abnormal, and the multi-dimensional impact data of printing production is optimized and controlled.

[0062] A database mapping dimensional position error to equipment and environment data is established based on the input and output of a decision tree model; a database mapping first color error to coating and environment data is established based on the input and output of a random forest regression model; and a database mapping second color error to color-coated sheet and environment data is established based on the input and output of a random forest regression model. The error tolerance thresholds stored in the databases are then obtained, including tolerance thresholds for dimensional error, position error, first color error, and second color error.

[0063] The multidimensional impact data combination of printing production to be evaluated was screened from the dimensional position error-equipment and environment data mapping database, the first color error-coating and environment data mapping database, and the second color error-color coated plate and environment data mapping database (all errors are required to be within the corresponding error allowable threshold; since the printing workshop environmental parameter data will be repeatedly acquired, the three sets of printing workshop environmental parameter data can be averaged). Based on the input and output of different models, dimensional position error-equipment and environment data, first color error-coating and environment data, and second color error-color coated plate and environment data mapping databases were established respectively. This helps to clearly understand the relationship between each influencing factor and error, and provides data support for subsequent analysis.

[0064] Based on the multidimensional impact data combinations of each candidate printing production, the corresponding comprehensive error value dYx is obtained (input into the decision tree model, multiple linear regression model, and random forest regression model); the cost-benefit Xyb and implementation feasibility score Pf of each candidate printing production multidimensional impact data combination are determined (relevant evaluation criteria are established in advance, and a comprehensive score is given based on the operability of the candidate printing production multidimensional impact data combination in actual production, including the ease of equipment adjustment (e.g., whether large-scale equipment modification is required), the technical requirements of environmental control (whether existing equipment can meet them), the complexity of coating formula improvement (whether special raw materials or processes are required), and the feasibility of adjusting the characteristics of color-coated plates (whether it affects the production cycle, etc.).

[0065] Determine the various evaluation coefficients (px);

[0066]

[0067] Where β1 is The weighting factors are β2 and β3 respectively. The weighting factor, e is the natural constant;

[0068] The combination of multidimensional impact data on printing production corresponding to the highest evaluation coefficient is used as the adjustment target to optimize and control the multidimensional impact data on printing production.

[0069] The evaluation coefficients are calculated using a set formula, and the data combination corresponding to the maximum evaluation coefficient is used as the adjustment target. By comprehensively considering error values, cost-effectiveness, and feasibility, the optimal parameter adjustment scheme can be selected to achieve precise optimization and control of multi-dimensional impact data on printing production, effectively improving printing accuracy while balancing quality, cost, and feasibility factors in production.

[0070] An electronic device includes: a processor; and a memory storing computer program instructions, which, when executed by the processor, cause the processor to perform the image recognition-based method for detecting the printing accuracy of color-coated steel plate patterns as described above.

[0071] A computer-readable storage medium for storing a program that, when executed by a processor, implements the image recognition-based method for detecting the printing accuracy of color-coated steel plate patterns as described above.

[0072] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0073] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0074] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0075] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0076] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0077] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for detecting the printing accuracy of color patterns on color-coated steel sheets based on image recognition, characterized in that, Includes the following steps: Obtain multi-dimensional impact data on printing production, and determine whether the printing environment for color-coated steel plate patterns is normal based on the multi-dimensional impact data generated during printing: If the printing environment for the color-coated steel plate is abnormal, printing shall be stopped. If the printing environment for the color and pattern of the color-coated steel plate is normal, then obtain the printed color-coated steel plate, extract the color-coated steel plate image for preprocessing, and obtain the preprocessed color-coated steel plate image. The image of the color-coated steel plate is processed to obtain the color pattern of the color-coated steel plate, and the printing accuracy error coefficient is obtained. The printing accuracy error threshold is determined based on multidimensional impact data of printing production, and the printing accuracy is evaluated by combining the printing accuracy error coefficient. Multidimensional impact data on printing production includes printing equipment operation data, printing workshop environmental parameter data, coating characteristic parameter data, and color-coated plate characteristic parameter data; Determining the printing accuracy error threshold based on multidimensional impact data of printing production includes the following steps: Using printing equipment operation data and printing workshop environmental parameter data as inputs to the decision tree model, the theoretical error of the size is obtained. and position theory error ; Using coating characteristic parameter data and printing workshop environmental parameter data as inputs to a multiple linear regression model, the first color theory error is obtained. ; Using the characteristic parameters of the color-coated steel sheet and the environmental parameters of the printing workshop as inputs to the random forest regression model, the theoretical error of the second color is obtained. ; Based on size theory error Position theory error First color theory error Second color theoretical error Obtain the printing accuracy error threshold : ； in, for Weighting factors for Weighting factors for Weighting factors.

2. The method for detecting the printing accuracy of color-coated steel plate patterns based on image recognition according to claim 1, characterized in that, Determining whether the printing environment for the color and pattern of color-coated steel sheets is normal based on multi-dimensional influence data generated from printing includes the following steps: The printing equipment operation data is used as input to the support vector machine model to obtain the printing equipment operation status. If the printing equipment operation status is abnormal, the printing environment of the color plate pattern is abnormal. If the printing equipment is operating normally, the environmental parameter data of the printing workshop will be used as the input of the environmental factor judgment model built on the convolutional neural network to obtain the environmental status of the printing workshop. If the environmental status of the printing workshop is abnormal, the printing environment of the color-coated plate pattern is abnormal. If the printing workshop environment is normal, the coating characteristic parameter data and the color-coated plate characteristic parameter data are used as input to the random forest model to obtain the coating characteristics and color-coated plate characteristic status. If the coating characteristics and color-coated plate characteristic status are abnormal, then the color-coated plate pattern printing environment is abnormal. If the characteristics of the coating and the color-coated sheet are normal, then the printing environment for the color-coated sheet patterns is normal.

3. The method for detecting the printing accuracy of color patterns on color-coated steel sheets based on image recognition according to claim 1, characterized in that, The process of pattern processing on the image of the color-coated steel sheet to obtain the color pattern of the color-coated steel sheet includes the following steps: Edge detection is performed on the color-coated steel sheet image to obtain the edge contour image; Corner detection algorithms are used to detect corners in edge contour images to obtain a set of corner coordinates. Threshold segmentation is performed based on edge contour images and corner coordinate sets to obtain color patterns on color-coated steel sheets.

4. The method for detecting the printing accuracy of color patterns on color-coated steel sheets based on image recognition according to claim 3, characterized in that, Edge detection is performed on the color-coated steel sheet image to obtain the edge contour image, including the following steps: After smoothing the image of the color-coated plate, a gradient image is obtained based on the Sobel operator; Non-maximum suppression is applied to the gradient image to obtain a thinned edge image; Obtain the initial dual-threshold detection interval stored in the database; Retrieve the printing workshop environment-double threshold correction coefficient mapping database stored in the database, and retrieve the corresponding double threshold correction coefficient based on the current printing workshop environment parameter data; The initial double threshold detection interval is corrected based on the double threshold correction coefficient to obtain the corrected double threshold detection interval; Obtain the pixel values ​​of each pixel in the thinned edge image: If the pixel value of a certain pixel is greater than the maximum value in the corrected double threshold detection interval, then the pixel is identified as an edge pixel. If the pixel value of a certain pixel is less than the minimum value in the corrected double threshold detection interval, then the pixel is identified as a non-edge pixel. If the pixel value of a certain pixel is within the corrected double threshold detection range, then it is determined whether the pixel is connected to an edge pixel. If it is connected, it is identified as an edge pixel; if it is not connected, it is identified as a non-edge pixel. Connect all the edge pixels to obtain the edge contour image.

5. The method for detecting the printing accuracy of color-coated steel plate patterns based on image recognition according to claim 3, characterized in that, Threshold segmentation based on edge contour images and corner coordinate sets is performed to obtain the color pattern of the color-coated steel plate, including the following steps: Obtain the pixel grayscale values ​​of all pixels within the edge contour image, plot the grayscale value distribution, determine the most concentrated grayscale interval, and define the most concentrated grayscale interval as the initial threshold segmentation range. Iterate through the pixel grayscale values ​​of the neighborhood pixels of each corner point in the corner coordinate set, and obtain the mean and standard deviation. If the mean is lower than the mean of the pixel grayscale values ​​of all pixels in the edge contour image and the standard deviation is greater than the set standard deviation threshold, then obtain the contaminant data and correct the initial threshold segmentation range based on the contaminant data to obtain the corrected threshold segmentation range; otherwise, do not correct the initial threshold segmentation range. If the grayscale value of a pixel in the color-coated plate image is within the corrected threshold segmentation range or the uncorrected initial threshold segmentation range, then the pixel is recorded as the pattern area; otherwise, it is recorded as the background area, thus obtaining the color pattern of the color-coated plate.

6. The method for detecting the printing accuracy of color patterns on color-coated steel sheets based on image recognition according to claim 1, characterized in that, Obtaining the printing accuracy error coefficient includes the following steps: Retrieve the standard pattern of color-coated steel sheet stored in the database, and align the color-coated steel sheet pattern with the standard pattern of color-coated steel sheet. Obtaining dimensional error Position error and color error ; Based on the error accuracy evaluation formula, the printing accuracy error coefficient is obtained. : ； in, for Weighting factors for Weighting factors for Weighting factors.

7. The method for detecting the printing accuracy of color-coated steel plate patterns based on image recognition according to claim 1, characterized in that, The evaluation of printing accuracy, incorporating a printing accuracy error coefficient, includes the following steps: If the printing accuracy error coefficient is less than the printing accuracy error threshold, the accuracy test result is normal. If the printing accuracy error coefficient is not less than the printing accuracy error threshold, the accuracy detection result is abnormal, and the multi-dimensional impact data of printing production will be optimized and controlled.

8. The method for detecting the printing accuracy of color patterns on color-coated steel sheets based on image recognition according to claim 7, characterized in that, Optimizing and controlling the multidimensional impact data of printing production includes the following steps: A database mapping size and position error between equipment and environment data is established based on the input and output of the decision tree model; a database mapping first color error between coating and environment data is established based on the input and output of the random forest regression model; and a database mapping second color error between color-coated steel sheet and environment data is established based on the input and output of the random forest regression model. Obtain the error tolerance thresholds stored in the database, including the size error tolerance threshold, the position error tolerance threshold, the first color error tolerance threshold, and the second color error tolerance threshold; The data sets of multidimensional impact data on printing production to be evaluated are selected from the dimensional position error-equipment and environment data mapping database, the first color error-coating and environment data mapping database, and the second color error-color coated plate and environment data mapping database. Based on the combination of multi-dimensional impact data on each candidate printing production, the corresponding comprehensive error values ​​for each candidate are obtained. ; Determine the cost-effectiveness of each combination of multidimensional impact data on print production to be evaluated. Feasibility assessment ; Determine each evaluation coefficient ; ； in, for Weighting factors for Weighting factors for The weighting factor, e is the natural constant; The combination of multidimensional impact data on printing production corresponding to the highest evaluation coefficient is used as the adjustment target to optimize and control the multidimensional impact data on printing production.

Images