Automatic intelligent printing quality detection method and system

Abstract

The invention relates to the technical field of printing quality detection, and discloses an automatic intelligent printing quality detection method and system, and the method comprises the steps: obtaining the printing design data of a target printing object, analyzing the display effect characteristics of the target printing object based on the design data, and generating an image detection scheme according to the display characteristics; according to the scheme, detection image data of a target printing object is collected, a pre-trained machine learning model is used for analyzing a detection image, a quality assessment report is generated, and the quality of the printing object is assessed, the printing quality is accurately detected in an automatic mode, errors caused by manual detection are avoided, and the detection efficiency is improved. The detection efficiency and consistency are improved, the printing defects and the quality problem can be accurately recognized by combining the printing design data and the machine learning model, and the problem that in the prior art, the efficiency of manual printing quality detection is low is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of printing quality detection, and in particular to an automated intelligent printing quality detection method and system. Background Art

[0002] In traditional printing quality inspection, manual judgment is easily affected by subjective factors and has low efficiency. As the modern printing industry continues to increase its requirements for production efficiency and product quality, traditional manual inspection methods can no longer meet the growing demand. Summary of the invention

[0003] The purpose of the present invention is to provide an automated intelligent printing quality detection method and system, aiming to solve the problem of low efficiency of manual printing quality detection in the prior art.

[0004] The present invention is implemented in this way. In a first aspect, the present invention provides an automated intelligent printing quality detection method, comprising: Acquire printing design data of a target printing object, and perform characteristic analysis of a display effect of a printed product on the target printing object according to the printing design data to obtain printing display characteristics of the target printing object; Performing an image detection scheme analysis on the target printed object according to the printed display characteristics to obtain an image data detection scheme for the target printed object; Collecting image data of the target printing object according to the image data detection scheme to obtain detection image data of the target printing object; The detection image data is analyzed and processed according to a pre-trained machine learning model to obtain a quality assessment report of the target printed object.

[0005] In a second aspect, the present invention provides an automated intelligent printing quality detection system, which is used to implement an automated intelligent printing quality detection method according to any one of the first aspects, including: A display analysis module, used to obtain printing design data of a target printing object, and perform a characteristic analysis of a display effect of a printed product on the target printing object according to the printing design data, so as to obtain printing display characteristics of the target printing object; A scheme analysis module, used for analyzing the image detection scheme of the target printed object according to the printing display characteristics, so as to obtain the image data detection scheme of the target printed object; A detection execution module, used for collecting image data of the target printing object according to the image data detection scheme to obtain detection image data of the target printing object; The quality assessment module is used to analyze and process the detection image data according to a pre-trained machine learning model to obtain a quality assessment report of the target printed object.

[0006] The present invention provides an automated intelligent printing quality detection method, which has the following beneficial effects: The present invention obtains printing design data of a target printing object, analyzes display effect characteristics of the target printing object based on the design data, generates an image detection scheme according to the display characteristics, collects detection image data of the target printing object according to the scheme, analyzes the detection image using a pre-trained machine learning model, generates a quality assessment report, and evaluates the quality of the printed object. The method accurately detects printing quality in an automated manner, avoids errors in manual detection, and improves detection efficiency and consistency. By combining printing design data and machine learning models, printing defects and quality problems can be accurately identified, solving the problem of low efficiency in manual detection of printing quality in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of the steps of an automated intelligent printing quality detection method provided by an embodiment of the present invention; Figure 2 It is a structural schematic diagram of an automated intelligent printing quality detection system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0008] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0009] The implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0010] Reference Figure 1 , Figure 2 As shown, a preferred embodiment of the present invention is provided.

[0011] In a first aspect, the present invention provides an automated intelligent printing quality detection method, comprising: S1: Acquire printing design data of a target printing object, and perform a characteristic analysis of a display effect of a printed product on the target printing object according to the printing design data to obtain printing display characteristics of the target printing object; S2: analyzing the image detection scheme of the target printed object according to the printing display characteristics to obtain an image data detection scheme of the target printed object; S3: collecting image data of the target printing object according to the image data detection scheme to obtain detection image data of the target printing object; S4: Analyze and process the detection image data according to a pre-trained machine learning model to obtain a quality assessment report of the target printing object.

[0012] Specifically, in step S1 of the embodiment provided by the present invention, the printing design data of the target printing object is obtained. The target printing object is a finished product to be prepared through printing work and is also an object to be prepared for printing quality inspection work.

[0013] More specifically, the printing contents and printing forms of different target printing objects are different, so different detection schemes need to be used for different target printing objects to achieve the best detection effect.

[0014] It should be noted that, in the technical solution provided by the present invention, image data to be detected are obtained by capturing images of the target printed object, and these image data are identified to perform quality detection on the target printed object. Under different optical environments, the image content displayed by the target printed object is different, that is, after applying different optical environments to the target printed object, image capture is performed on the target printed object in different optical environments, and the captured images are different. The differential images displayed by the target printed object will be more conducive to the detection of certain types of defects, that is, certain defects that are difficult to display on the target printed object under normal optical environments will be easier to identify under specific optical environments.

[0015] Furthermore, the printed content and printing form on the target printed object are different, and the target printed objects with different printed content and printing form will display different effects in different optical environments. This means that different target printed objects will be more likely to expose printing quality defects in different optical environments. In order to obtain this information, it is necessary to obtain the printing design data of the target printed object, and then perform feature analysis on the display effect of the printed product of the target printed object based on the printing design data to obtain the printing display characteristics of the target printed object, that is, what kind of display effect the target printed object will display in different optical environments, so as to determine what optical environment filter to use to detect the target printed object in subsequent steps.

[0016] It should be noted that the optical environment filter refers to the specific optical environment applied to the target printed object when capturing images of the target printed object. The specific optical environment includes lighting effects such as light intensity, light frequency, and light color.

[0017] Specifically, in step S2 of the embodiment provided by the present invention, an image detection scheme is analyzed for the target printed object according to the printing display characteristics to obtain an image data detection scheme for the target printed object.

[0018] More specifically, the printing display characteristics of the target printed object feedback the display information of the target printed object under different optical environment filters. By analyzing the display information, the detection adaptability of the display information to various defect types can be obtained, that is, the display information of the target printed object under the optical environment filter for the detection effect of various defect types of the target printed object.

[0019] More specifically, the printed content at different locations on the target printed object is different, so the most suitable optical environment filters for different locations are different. Based on this consideration, different optical environment filters need to be applied to the target printed object multiple times to perform comprehensive and accurate defect detection on various locations of the target printed object.

[0020] More specifically, by analyzing the printing display characteristics of the target printed object, optical environment filters that can best display various defects of the target printed object are obtained, and these optical environment filters are converted into image data detection schemes for the target printed object, that is, image data of the target printed object is collected under these formulated optical environment filters to obtain detection image data to be detected.

[0021] More specifically, since each optical environment filter is most suitable for detecting different defect types, in order to cover all aspects, it is necessary to combine multiple optical environment filters to obtain multiple optical environment filters with the most comprehensive detection as an image data detection solution.

[0022] Specifically, in step S3 of the embodiment provided by the present invention, different optical environment filters are applied to the target printed object in sequence according to the image data detection scheme, and then image data of the target printed object under different optical environment filters is collected to obtain detection image data of the target printed object under different optical environment filters.

[0023] It is understandable that the detection image data of the target printed object under different optical environment filters have different adaptability to the detection of various types of defects. For example, the brightness of a certain type of color printing is difficult to identify within a certain range under normal optical environment. However, by adjusting the brightness and color of the optical environment, the brightness of this type of color printing can be more easily reflected in the image data. Therefore, multiple inspections are required under different optical environment filters to ensure the comprehensiveness and accuracy of quality inspection.

[0024] Specifically, in step S4 of the embodiment provided by the present invention, after obtaining the detection image data, the detection image data is analyzed and processed using a pre-trained machine learning model to obtain a quality assessment report of the target printed object.

[0025] It is understandable that the use of machine learning models and image analysis technology can automatically detect problems in the printing process, avoid subjective bias in manual inspection, improve inspection efficiency and accuracy, and reduce manual operation and inspection time through automated quality analysis and report generation, while reducing error rates and saving costs.

[0026] The present invention provides an automated intelligent printing quality detection method, which has the following beneficial effects: The present invention obtains printing design data of a target printing object, analyzes display effect characteristics of the target printing object based on the design data, generates an image detection scheme according to the display characteristics, collects detection image data of the target printing object according to the scheme, analyzes the detection image using a pre-trained machine learning model, generates a quality assessment report, and evaluates the quality of the printed object. The method accurately detects printing quality in an automated manner, avoids errors in manual detection, and improves detection efficiency and consistency. By combining printing design data and machine learning models, printing defects and quality problems can be accurately identified, solving the problem of low efficiency in manual detection of printing quality in the prior art.

[0027] Preferably, the step of acquiring printing design data of a target printing object, and performing characteristic analysis of a display effect of a printed finished product on the target printing object according to the printing design data to obtain printing display characteristics of the target printing object includes: S11: Acquire printing design data of a target printing object; wherein the target printing object includes a substrate portion and a printing portion, and the printing design data is used to describe a specific form of the printing portion pre-planned and displayed on the substrate portion; S12: performing data analysis on the shape dimension and the color dimension on the printing design data to obtain shape data features and color data features of the target printing object; S13: performing simulation analysis on the display effects of the morphological data features and the color data features of the target printed object according to several preset optical environment filters, so as to obtain the morphological display effect maps and the color display effect maps of the target printed object corresponding to various optical environment filters; S14: combining data of the morphology display effect map and the color display effect map corresponding to various optical environment filters of the target printing object to obtain the printing display characteristics of the target printing object.

[0028] Specifically, the printing design data of the target printing object is obtained. These data describe the specific form of the printed part on the target printing object (including the substrate and the printed part), such as the design pattern, font, color and texture, etc. The design data usually comes from CAD files, vector graphics files (such as AI, SVG files) or PDF and other formats. The design data is extracted through appropriate file parsing tools, and the predetermined position, shape, color and other information of the printed part on the substrate are structured and stored to ensure that the data can reflect the detailed design of the printed object.

[0029] More specifically, the acquired design data is analyzed in terms of morphological dimension and color dimension to extract the morphological data features and color data features of the target printing object. The morphological dimension includes pattern, size, proportion, etc., and the color dimension includes hue, saturation, brightness, etc. The design data is analyzed using image analysis and pattern recognition technology. For the morphological dimension, the outline, shape, size and other features of the printing design can be extracted through geometric analysis (such as graphic segmentation, edge detection, etc.); for the color dimension, the specific characteristics of the color can be obtained through color space conversion (such as RGB to CMYK or Lab color space conversion).

[0030] More specifically, the morphological data characteristics and color data characteristics of the target printed object are simulated and analyzed according to a preset plurality of optical environment filters. By simulating different optical conditions and observing the changes in the printing effect, the performance of the target printed object in different environments is evaluated.

[0031] More specifically, by simulating different optical environment filters through illumination models (such as the Phong model or the Blinn-Phong model), the rendering technology in computer graphics can be used to simulate the surface reflection effects (such as highlights, diffuse reflection, refraction, etc.) of printed materials under different lighting conditions, and apply them to the morphology and color characteristics of the target printed object to generate a corresponding display effect map.

[0032] More specifically, based on the results of simulation analysis, the morphology display effect map and color display effect map of the target printed object under various optical environment filters are generated. These maps show the morphology and color performance of the printed object under different optical environments, which can help designers evaluate whether the design meets the expected effect. Image processing and data visualization technology (such as heat map, color difference map, etc.) are used to generate morphology display effect map and color display effect map. These maps can show the detailed changes under different optical conditions (such as surface gloss, color change, contrast change, etc.).

[0033] More specifically, the morphology display effect map generated under different optical environment filters is combined with the color display effect map to comprehensively evaluate the overall display characteristics of the target printed object. This process helps to identify potential problems in the design and evaluate the performance of printed materials in various environments. The morphology and color characteristics can be combined through multidimensional data fusion technology (such as principal component analysis PCA, weighted average, etc.). In this way, comprehensive printing display feature data can be generated and output as charts or reports to help designers quickly understand the display effects of target printed objects in different environments.

[0034] More specifically, by combining the atlas and data, the printing display characteristics of the target printing object are finally obtained. These characteristics can reflect the morphological changes and color performance of the printed product in different optical environments, providing a scientific basis for subsequent quality inspection.

[0035] Preferably, the step of simulating and analyzing the display effects of the morphological data features and the color data features of the target printed object according to several preset optical environment filters to obtain the morphological display effect maps and the color display effect maps of the target printed object corresponding to various optical environment filters includes: S131: performing subdivision positioning processing on the shape data features and the color data features of the target printing object based on the grid coordinate system to obtain display versions of the shape data features and the color display features corresponding to the grid coordinate system, and marking the display versions of the shape data features and the color display features corresponding to the grid coordinate system as shape grid feature distribution and color grid feature distribution respectively; S132: simulating the environmental correction effect of the shape grid feature distribution and the color grid feature distribution respectively according to the preset optical environment filter, and assigning grid feature parameters to the shape grid feature distribution and the color grid feature distribution respectively according to the result of the environmental correction effect simulation, so as to obtain the shape grid parameter distribution and the color grid parameter distribution of the target printing object corresponding to the optical environment filter; S133: Repeat the above steps for various preset optical environment filters to obtain the shape grid parameter distribution and color grid parameter distribution of the target printed object corresponding to the various optical environment filters; S134: performing a comprehensive evaluation of grid-independent display effects and grid-adjacent display effects on the morphology grid parameter distributions and color grid parameter distributions of the target printed object corresponding to the various optical environment filters, so as to obtain morphology display effect levels and color display effect levels of the target printed object corresponding to the various optical environment filters; S135: The morphology display effect levels of the target printing object corresponding to the various optical environment filters together constitute a morphology display effect map, and the color display effect levels of the target printing object corresponding to the various optical environment filters together constitute a color display effect map.

[0036] Specifically, the morphological data features and color display features of the target printing object are subdivided and located on the grid coordinate system, which means mapping the object's morphological and color data into a predetermined grid. A two-dimensional or three-dimensional grid coordinate system is usually used to represent the features of each local area. According to the size and complexity of the target printing object, a suitable grid coordinate system is created, and the shape, texture, pattern and other features and color information of the target printing object are mapped into the grid to generate a morphological grid feature distribution and a color grid feature distribution. Each grid node will contain the morphological and color data of the target object in the area.

[0037] More specifically, the morphology grid feature distribution and the color grid feature distribution are marked separately through the gridded subdivision data to ensure that the morphology and color information are stored and managed according to the grid distribution. The morphology grid feature distribution will contain the geometry, texture, surface smoothness and other data of each grid unit; the color grid feature distribution contains the color value, brightness, saturation and other information of each grid unit.

[0038] More specifically, the environmental correction effect is simulated on the morphology grid feature distribution and the color grid feature distribution according to the preset optical environment filter. This step aims to simulate the changes in morphology and color in different grid units under different optical environments. Optical simulation technology (such as rendering technology based on ray tracing) is used to simulate the environmental correction effect. The environmental correction takes into account the influence of different lighting, reflection, refraction and other factors on the morphology and color of each grid unit. This will calculate the new value of the morphology and color of each grid unit according to the type of filter.

[0039] More specifically, according to the results of the environmental correction effect simulation, the parameters of the grid features are assigned to the morphology grid feature distribution and the color grid feature distribution respectively. This means that new morphology and color parameters are assigned to each grid unit according to the results of different optical environment corrections. The morphology and color feature parameters of each grid unit will be assigned new values ​​based on the data after environmental simulation. These values ​​include changes in morphology (such as surface gloss, unevenness, etc.) and color changes (such as color difference, saturation changes, etc.).

[0040] More specifically, the above steps are repeated to simulate the correction effect of the target printed object under multiple preset optical environment filters, and the morphology grid parameter distribution and the color grid parameter distribution are obtained respectively. This will generate different morphology and color effects for each environment filter. Different lighting conditions, reflection modes, etc. are configured for each optical environment filter. The above process is repeated using a rendering engine or an optical simulation algorithm to generate a new set of morphology and color parameter distributions under each filter.

[0041] More specifically, a comprehensive evaluation is conducted on the morphology grid parameter distribution and color grid parameter distribution under each optical environment filter, taking into account the independent display effect of the grid unit and the interactive effect of the adjacent grids. Image processing techniques such as high-pass filtering, edge detection, adjacency analysis, etc. are used to independently evaluate the morphology and color changes of each grid unit. At the same time, the transition effects and mutual influences between adjacent grids are also considered. The comprehensive evaluation will take into account both local details and overall effects.

[0042] More specifically, based on the results of the comprehensive evaluation, a morphology display effect level and a color display effect level are generated for each optical environment filter. One level represents the display performance of the target printed object under an optical environment filter. The morphology display effect levels and color display effect levels under different optical environment filters are merged to generate the final morphology display effect atlas and color display effect atlas. Through data visualization technology (such as heat maps, color scale maps, stereograms, etc.), the display effect levels under different environment filters are merged to generate morphology display effect atlases and color display effect atlases. These atlases can display the morphology and color changes in each environment.

[0043] Preferably, the step of analyzing the image detection scheme of the target printed object according to the printing display feature to obtain the image data detection scheme of the target printed object includes: S21: performing an optical environment filter evaluation on the morphology display effect map and the color display effect map in the printing display feature and applying them to the target printing object, so as to obtain a plurality of optical environment filters to be used in the detection work; S22: converting a test execution scheme for a plurality of optical environment filters to be used in the detection work to obtain a filter execution scheme for applying the optical environment filter to the target printed object; wherein the filter execution scheme includes filter execution parameters and filter execution time; S23: constructing a corresponding image acquisition execution scheme for the filter execution time in each of the filter execution schemes to obtain an image acquisition execution scheme corresponding to each of the filter execution schemes; S24: Each of the filter execution schemes and the image acquisition execution scheme corresponding to each of the filter execution schemes together constitute the image data detection scheme.

[0044] Specifically, the optical environment filters of the target printed object are evaluated using the morphology display effect atlas and the color display effect atlas. The purpose of this step is to identify the optical environment conditions related to the display effect of the target printed object in actual application, and to screen out the most suitable optical environment filters for detection by comparing the morphology and color performance of the target printed object in different optical environments. By simulating the display effects under different optical environment filters, the morphology and color characteristics of the target printed object are evaluated, thereby selecting the best evaluated optical environment filter set, which will be used for subsequent image detection work.

[0045] More specifically, a test execution plan is formulated for the best optical environment filter evaluated in the previous step, and converted into a specific filter execution plan. These filter execution plans will include the execution parameters of each optical environment filter (how to apply the optical environment filter to the target printed object) and the execution time (the action time of each filter). According to the characteristics of different optical environment filters, they are converted into specific filter execution plans. In this process, the execution parameters of the filter need to be clearly set, and the execution time of each filter needs to be determined to ensure that the test effect matches the display effect of the target printed object.

[0046] More specifically, according to the execution time of each optical environment filter, a corresponding image acquisition execution plan is constructed. The image acquisition execution plan will be time-planned according to the filter execution time to ensure that the image acquisition process can acquire the required image data at the correct time. According to the filter execution time, the image acquisition time interval, image resolution, acquisition frequency and other parameters are formulated. Each filter execution plan may correspond to one or more image acquisition time points. The construction of the acquisition plan will take into account the performance changes of image data under different filters. For example, for some filters, long exposure may be required to capture changes in lighting, while for other filters, only short-term acquisition may be required.

[0047] More specifically, the execution plan of each optical environment filter and the corresponding image acquisition execution plan are combined to form a final image data detection plan. This plan will include specific operating steps for applying filters to target printed objects under different optical environments, as well as corresponding image acquisition plans under each filter. The combination of each filter execution plan and image acquisition execution plan forms a complete detection process. This detection process will include parameter settings, execution time, acquisition plans, etc. for each optical environment filter. During the entire process, an automated detection system (such as computer vision, image processing platform) can be used to execute the operations in the plan to ensure that image data can be acquired efficiently and accurately.

[0048] Preferably, the step of evaluating the optical environment filter applied to the target printed object on the morphology display effect map and the color display effect map in the printed display feature to obtain a plurality of optical environment filters to be used in the detection work comprises: S211: performing a detection adaptability analysis of the corresponding quality defect types on each of the optical environment filters according to the morphology display effect map and the color display effect map in the printing display feature, so as to obtain the detection adaptability characteristic distribution of the various types of optical environment filters fed back by the morphology display effect map and the color display effect map in the printing display feature; wherein the detection adaptability characteristic distribution is used to describe the detection effect of the optical environment filter for various types of quality defects; S212: Retrieving various quality defect types and corresponding detection weights of the target printing object from a preset database, and using the various quality defect types and corresponding detection weights as detection purposes; S213: analyzing the achievement status of the detection purpose according to the detection adaptability characteristic distribution of each type of optical environment filter, so as to obtain the detection purpose achievement characteristics of each type of optical environment filter corresponding to the detection purpose; S214: performing filter combinations for various optical environment filters according to the detection purpose achievement characteristics of various optical environment filters, and performing multi-dimensional analysis on the filter combinations in terms of detection purpose achievement status, filter execution resource consumption, and filter detection error self-correction, so as to obtain a number of filter combinations and combination evaluation values; S215: Using each optical environment filter in the filter combination with the best combination evaluation value as a number of optical environment filters to be prepared for the detection work.

[0049] Specifically, the adaptability of each optical environment filter is analyzed separately in combination with the morphology display effect map and the color display effect map. These display maps will feedback the performance of the target printed object in different optical environments, including morphology and color characteristics. Based on these characteristics, the adaptability of each optical environment filter to the detection of various quality defects (such as color difference, uneven printing, morphological defects, etc.) is evaluated. Computer vision and image processing algorithms are used to analyze the morphology and color maps, and according to the characteristics of each filter, the detection effects under different quality defect types are simulated. The adaptability feature distribution of the filter is established through machine learning or statistical methods, and the performance characteristics of each filter for different defect types are generated.

[0050] More specifically, based on the detection adaptability feature distribution of the optical environment filter obtained from the first step, combined with the target detection task, that is, the preset quality defect type and its corresponding detection weight, the performance of each filter in actual detection is analyzed. Here, the weight of the quality defect type reflects the importance of various defects in the actual production process, thereby providing a basis for subsequent filter evaluation. By matching the adaptability characteristics of each optical environment filter with the preset quality defect type and its weight, a multi-dimensional target achievement analysis is performed. This process calculates the degree of matching between the filter characteristics and the target to generate a filter "target achievement status" report to measure the effectiveness of each filter in the detection task.

[0051] More specifically, according to the "goal achievement status" of the second step and the detection purpose achievement characteristics of various filters, a filter combination analysis is performed. In this step, various optical environment filters are combined to evaluate multi-dimensional factors such as detection effects, resource consumption, detection errors and error self-correction of different combinations. Finally, the filter combination with the best combination evaluation value is selected, and different filter combinations are evaluated using multi-objective optimization algorithms (such as genetic algorithms, simulated annealing algorithms, etc.). During the evaluation process, the following factors are considered: Detection purpose achievement status: the ability of the filter combination to achieve the preset goal (i.e., effectively detect various defects), filter execution resource consumption: including filter execution time, computing resources, hardware requirements, etc., filter detection error and self-correction: the error of each filter combination is evaluated, and those filter combinations that can automatically correct the detection error are selected. These factors are weighted to generate a comprehensive evaluation value to select the optimal filter combination.

[0052] More specifically, the filter combination evaluation results obtained through multi-dimensional analysis will select the optical environment filter combination with the best comprehensive evaluation value. These filters will be used as optical environment filters to be prepared for detection. After comprehensive consideration of various evaluation dimensions, the best combination will be selected and applied to actual detection. The selection of filter combinations not only ensures the optimization of detection effects, but also improves the resource utilization efficiency during the detection process and minimizes errors and unnecessary waste of resources.

[0053] Preferably, the step of collecting image data of the target printing object according to the image data detection scheme to obtain the detection image data of the target printing object includes: S31: applying optical environment filters to the target printing object in sequence according to each of the filter execution schemes in the image data detection scheme, so that the target printing object is placed in each of the specified optical environment filters in sequence; S32: performing image acquisition on the target printed object in each specified optical environment filter according to the image acquisition execution scheme in the image data detection scheme to obtain detection image data of the target printed object; wherein the detection image data includes detection images under each optical environment filter.

[0054] Specifically, according to the filter execution scheme in the image data detection scheme, each optical environment filter is applied to the target printed object in turn. Each optical environment filter corresponds to a specific detection environment, which may include different optical properties such as light, color temperature, contrast, etc. These filters can simulate different environmental conditions, with the purpose of testing the quality of the target printed object under various optical environments. Different optical environment filters are applied to the target printed object in turn using an optical filter device or a combination of an adjustable light source, a camera and an optical component. These filters simulate different detection conditions by changing light properties (such as brightness, hue, contrast, etc.) to ensure comprehensive and multi-dimensional detection of the target printed object.

[0055] More specifically, according to the image acquisition execution scheme in the image data detection scheme, image acquisition is performed on the performance of the target printed object under each optical environment filter to obtain the detection image data under each filter condition. Each detection image data reflects the printing quality characteristics of the target printed object under the specific optical environment. A high-precision camera or image acquisition device is used to perform image acquisition under different optical environment filters. This process usually involves using automated equipment to scan and photograph the target object to ensure that the image data under each optical environment filter is accurately recorded. It may involve multiple imaging technologies, such as visible light imaging, infrared imaging, ultraviolet imaging, etc., to capture different quality characteristics.

[0056] More specifically, through the above-mentioned acquisition process, the detection image data of the target printed object under different optical environment filters are obtained. These detection image data constitute a multi-level image data set. Each image reflects the performance of the target printed object under a specific optical environment. Multiple image data are generated and stored. Each image corresponds to the quality performance of the target printed object under a specific filter. It can be saved in a high-resolution image format (such as TIFF, PNG, JPEG), and the relevant optical environment filter information and image acquisition metadata are attached for subsequent analysis and processing.

[0057] Preferably, the pre-training step of the machine learning model includes: S41: Collecting model learning data of the machine learning model; wherein the model learning data includes detection image data and corresponding image evaluation annotations, the image evaluation annotations are used to annotate the detection image data with quality evaluation results, and the image evaluation annotations correspond to the quality evaluation report; S42: Divide the model learning data into sets to obtain a training set, a validation set, and a test set; S43: Construct a machine learning model consisting of an input layer, a multi-layer recurrent layer, and an output layer; S44: Substituting the training set into the machine learning model, and allowing the machine learning model to perform feature analysis on the data mapping relationship between the detection image data and the image evaluation annotation according to the training set, so as to obtain the data mapping relationship between the detection image data and the image evaluation annotation; S45: Substituting the verification set into the machine learning model, and allowing the machine learning model to verify the data mapping relationship according to the verification set to obtain the confidence of the data mapping relationship; S46: adjusting the model parameters of the machine learning model according to the confidence of the data mapping relationship to obtain a new data mapping relationship; S47: Substituting the test set into the machine learning model, allowing the machine learning model to test the data mapping relationship according to the test set, and obtaining the data fitting status of the data mapping relationship; S48: When the data fitting condition of the data mapping relationship meets the training standard, the machine learning model having the data mapping relationship is deployed to complete the training of the machine learning model; S49: When the data fitting condition of the data mapping relationship does not meet the training standard, the data mapping relationship is trained again until the data mapping relationship whose data fitting condition meets the training standard is obtained.

[0058] Specifically, it is necessary to collect model learning data suitable for training, which includes detection image data and corresponding image evaluation annotations. The image evaluation annotations are used to perform quality assessment on each detection image data and record the corresponding evaluation results. The image evaluation annotations may include information such as the type of defects in the image, the severity, and the printing quality grade. These data will be used as label data for supervised learning. Each image will be evaluated through manual annotation or automated tools (such as image processing technology, computer vision algorithms, etc.), and the quality annotation information will be attached to the image data to form a pair of training samples.

[0059] More specifically, the model learning data (including detection image data and image evaluation annotations) is divided into sets, usually into training set, validation set and test set: Training set: used to train the machine learning model and adjust the model's parameters; Validation set: used to monitor the performance of the model during training, adjust hyperparameters, and select the optimal model; Test set: used to evaluate the final performance of the model and verify whether the model has good generalization ability. The common division ratio is 70% for training set, 15% for validation set, and 15% for test set. The representativeness and uniformity of the data must be guaranteed during division to ensure that the data in each subset is representative.

[0060] More specifically, according to the complexity of the problem and the characteristics of the data, a suitable machine learning model is constructed. According to the description, the model structure is a neural network model consisting of an input layer, a multi-layer recurrent layer (such as LSTM, GRU, etc.) and an output layer. The input layer: accepts processed image data as input. The multi-layer recurrent layer: is used to learn the temporal information and spatial relationship in the image data, usually used for tasks such as image sequence and video analysis. The output layer: outputs the quality assessment annotation of the image according to the task requirements, such as classification labels or regression results. Choose a suitable framework (such as TensorFlow, PyTorch, etc.) to build the neural network model and configure appropriate hyperparameters such as the number of layers, activation functions, and loss functions.

[0061] More specifically, the training set data is substituted into the constructed machine learning model, and the model is made to learn based on the relationship between the image data in the training set and the image evaluation annotations. The model performs feature analysis and parameter updates on the data mapping relationship through the back propagation algorithm, and optimizes the model parameters through the gradient descent method or other optimization algorithms (such as Adam, RMSprop, etc.) to minimize the error between the predicted results and the actual annotations.

[0062] More specifically, the validation set is substituted into the model, and the model verifies the learned mapping relationship based on the validation set. The validation set is used to evaluate the model's confidence in the data mapping relationship, calculate the error or loss function value, and determine whether the model hyperparameters need to be adjusted. After each round of training, the validation set is used for prediction, and indicators such as accuracy, F1 score, and loss are calculated to evaluate the performance of the model. If the model does not perform well on the validation set, hyperparameters (such as learning rate, batch size, etc.) need to be adjusted.

[0063] More specifically, based on the confidence evaluation results on the validation set, the model parameters (such as network weights, learning rate, etc.) are adjusted to optimize the data mapping relationship. This process may include multiple training and validation cycles until the model reaches the expected performance level. Hyperparameter optimization algorithms (such as grid search, random search, or Bayesian optimization) are used to find the optimal parameter combination.

[0064] More specifically, the test set data is substituted into the model for testing to evaluate the fit of the data mapping relationship and check the performance of the model on unseen data. The role of the test set is to verify the generalization ability of the model and ensure that it performs well in real-world applications. During the testing phase, the test set is used to make model predictions and calculate common evaluation indicators (such as precision, recall, F1 score, AUC, etc.) to evaluate whether the model meets the expected standards.

[0065] More specifically, when the model's data fitting condition on the test set reaches the training standard, the model is considered to have been trained and has sufficient accuracy and can be deployed. The task of the deployment phase is to apply the trained model to actual scenarios for real-time prediction and decision-making.

[0066] More specifically, if it is found during the testing phase that the data fitting does not meet the training standards, the model needs to be retrained until a satisfactory fitting effect is achieved. Adjustments can be made by analyzing the training results (such as overfitting or underfitting), increasing the amount of data, changing the model structure, or using regularization techniques, etc., until the model performance meets the standards.

[0067] Reference Figure 2 As shown, in a second aspect, the present invention provides an automated intelligent printing quality detection system, which is used to implement an automated intelligent printing quality detection method as described in any one of the first aspects, including: A display analysis module, used to obtain printing design data of a target printing object, and perform a characteristic analysis of a display effect of a printed product on the target printing object according to the printing design data, so as to obtain printing display characteristics of the target printing object; A scheme analysis module, used for analyzing the image detection scheme of the target printed object according to the printing display characteristics, so as to obtain the image data detection scheme of the target printed object; A detection execution module, used for collecting image data of the target printing object according to the image data detection scheme to obtain detection image data of the target printing object; The quality assessment module is used to analyze and process the detection image data according to a pre-trained machine learning model to obtain a quality assessment report of the target printed object.

[0068] In this embodiment, for the specific implementation of each module in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0069] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An automated intelligent printing quality detection method, characterized in that: include: Acquire printing design data of a target printing object, and perform characteristic analysis of a display effect of a printed product on the target printing object according to the printing design data to obtain printing display characteristics of the target printing object; Performing an image detection scheme analysis on the target printed object according to the printed display characteristics to obtain an image data detection scheme for the target printed object; Collecting image data of the target printing object according to the image data detection scheme to obtain detection image data of the target printing object; The detection image data is analyzed and processed according to a pre-trained machine learning model to obtain a quality assessment report of the target printed object.

2. The automated intelligent printing quality detection method according to claim 1, characterized in that: The steps of acquiring printing design data of a target printing object, and performing characteristic analysis of a display effect of a printed finished product on the target printing object according to the printing design data to obtain printing display characteristics of the target printing object include: Acquire printing design data of a target printing object; wherein the target printing object includes a substrate portion and a printing portion, and the printing design data is used to describe a specific form of the printing portion pre-planned and displayed on the substrate portion; Performing data analysis on the shape dimension and the color dimension on the printing design data to obtain shape data features and color data features of the target printing object; According to a plurality of preset optical environment filters, a simulation analysis is performed on the display effect of the morphological data characteristics and the color data characteristics of the target printing object, so as to obtain a morphological display effect map and a color display effect map of the target printing object corresponding to various optical environment filters; The morphology display effect map and the color display effect map corresponding to various optical environment filters of the target printing object are combined to obtain the printing display characteristics of the target printing object.

3. The automated intelligent printing quality detection method according to claim 2, characterized in that: The steps of simulating and analyzing the display effects of the morphological data features and the color data features of the target printed object according to the preset optical environment filters to obtain the morphological display effect maps and the color display effect maps of the target printed object corresponding to the various optical environment filters include: Performing segmentation and positioning processing on the shape data features and the color data features of the target printing object based on the grid coordinate system, respectively, to obtain display versions of the shape data features and the color display features corresponding to the grid coordinate system, and marking the display versions of the shape data features and the color display features corresponding to the grid coordinate system as shape grid feature distribution and color grid feature distribution, respectively; According to a preset optical environment filter, the environmental correction effect of the morphology grid feature distribution and the color grid feature distribution are respectively simulated, and according to the result of the environmental correction effect simulation, the parameters of the grid features are respectively assigned to the morphology grid feature distribution and the color grid feature distribution, so as to obtain the morphology grid parameter distribution and the color grid parameter distribution of the target printing object corresponding to the optical environment filter; Repeat the above steps for various preset optical environment filters to obtain the shape grid parameter distribution and color grid parameter distribution of the target printed object corresponding to the various optical environment filters; Comprehensively evaluating the grid-independent display effect and the grid-adjacent display effect of the morphology grid parameter distribution and the color grid parameter distribution of the target printed object corresponding to the various optical environment filters, so as to obtain the morphology display effect level and the color display effect level of the target printed object corresponding to the various optical environment filters; The morphology display effect levels of the target printing object corresponding to the various optical environment filters together constitute a morphology display effect map, and the color display effect levels of the target printing object corresponding to the various optical environment filters together constitute a color display effect map.

4. The automated intelligent printing quality detection method according to claim 2, characterized in that: The step of analyzing the image detection scheme of the target printed object according to the printing display feature to obtain the image data detection scheme of the target printed object includes: Performing an optical environment filter evaluation on the morphology display effect map and the color display effect map in the printing display feature and applied to the target printing object, so as to obtain a plurality of optical environment filters to be prepared for the detection work; Performing test execution scheme conversion on a plurality of optical environment filters prepared to participate in the detection work, so as to obtain a filter execution scheme for applying the optical environment filter to the target printed object; wherein the filter execution scheme includes filter execution parameters and filter execution time; Constructing a corresponding image acquisition execution scheme for the filter execution time in each of the filter execution schemes to obtain an image acquisition execution scheme corresponding to each of the filter execution schemes; Each of the filter execution schemes and the image acquisition execution scheme corresponding to each of the filter execution schemes together constitute the image data detection scheme.

5. The automated intelligent printing quality detection method according to claim 4, characterized in that: The steps of evaluating the optical environment filter applied to the target printed object on the morphology display effect map and the color display effect map in the printed display feature to obtain a plurality of optical environment filters to be used in the detection work include: According to the morphology display effect map and the color display effect map in the printing display feature, the detection adaptability analysis of each of the optical environment filters corresponding to the quality defect type is performed to obtain the detection adaptability characteristic distribution of various types of optical environment filters fed back by the morphology display effect map and the color display effect map in the printing display feature; wherein the detection adaptability characteristic distribution is used to describe the detection effect of the optical environment filter for various types of quality defects; Retrieving various quality defect types and corresponding detection weights of the target printing object from a preset database, and using the various quality defect types and corresponding detection weights as detection purposes; Analyzing the achievement of the detection purpose according to the detection adaptability characteristic distribution of various optical environment filters to obtain the detection purpose achievement characteristics of various optical environment filters corresponding to the detection purpose; According to the detection purpose achievement characteristics of various optical environment filters, various optical environment filters are combined, and the filter combinations are analyzed in multiple dimensions, including the detection purpose achievement status, filter execution resource consumption, and filter detection error self-correction, so as to obtain several filter combinations and combination evaluation values; Each optical environment filter in the filter combination with the best combination evaluation value is used as a number of optical environment filters prepared to participate in the detection work.

6. The automated intelligent printing quality detection method according to claim 4, characterized in that: The step of collecting image data of the target printing object according to the image data detection scheme to obtain the detection image data of the target printing object includes: Applying optical environment filters to the target printed object in sequence according to each of the filter execution schemes in the image data detection scheme, so that the target printed object is placed in each of the specified optical environment filters in sequence; According to the image acquisition execution scheme in the image data detection scheme, image acquisition is performed on the target printed object in each specified optical environment filter to obtain the detection image data of the target printed object; wherein the detection image data includes the detection images under each optical environment filter.

7. The automated intelligent printing quality detection method according to claim 1, characterized in that: The pre-training steps for a machine learning model include: Collecting model learning data of a machine learning model; wherein the model learning data includes detection image data and corresponding image evaluation annotations, the image evaluation annotations are used to annotate the detection image data with quality evaluation results, and the image evaluation annotations correspond to the quality evaluation report; Dividing the model learning data into sets to obtain a training set, a validation set, and a test set; Build a machine learning model consisting of an input layer, multiple recurrent layers, and an output layer; Substituting the training set into the machine learning model, allowing the machine learning model to perform feature analysis on the data mapping relationship between the detection image data and the image evaluation annotation according to the training set, so as to obtain the data mapping relationship between the detection image data and the image evaluation annotation; Substituting the verification set into the machine learning model, and allowing the machine learning model to verify the data mapping relationship according to the verification set to obtain the confidence of the data mapping relationship; Adjusting the model parameters of the machine learning model according to the confidence of the data mapping relationship to obtain a new data mapping relationship; Substituting the test set into the machine learning model, allowing the machine learning model to test the data mapping relationship according to the test set to obtain the data fitting status of the data mapping relationship; When the data fitting status of the data mapping relationship meets the training standard, the machine learning model with the data mapping relationship is deployed to complete the training of the machine learning model; When the data fitting condition of the data mapping relationship does not meet the training standard, the data mapping relationship is trained again until the data mapping relationship whose data fitting condition meets the training standard is obtained.

8. An automated intelligent printing quality inspection system, characterized in that: An automated intelligent printing quality detection method for implementing any one of claims 1 to 7, comprising: A display analysis module, used to obtain printing design data of a target printing object, and perform a characteristic analysis of a display effect of a printed product on the target printing object according to the printing design data, so as to obtain printing display characteristics of the target printing object; A scheme analysis module, used for analyzing the image detection scheme of the target printed object according to the printing display characteristics, so as to obtain the image data detection scheme of the target printed object; A detection execution module, used for collecting image data of the target printing object according to the image data detection scheme to obtain detection image data of the target printing object; The quality assessment module is used to analyze and process the detection image data according to a pre-trained machine learning model to obtain a quality assessment report of the target printed object.

Images