Intelligent control method and system for color box production process

Abstract

The present application discloses an intelligent control method and system for color box production process, which uses deep learning-based image processing technology to process and analyze the image of the back-printed paper collected by the camera, so as to mine the traditional image feature expression and the deep-level image semantic feature expression, thereby combining the complementary feature information of the two to complete the description and characterization of the surface state of the back-printed paper. In this way, the monitoring and identification of printing defects can be realized.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent color box production, and in particular to an intelligent control method and system for a color box production process. Background Art

[0002] Cartons are common items in daily life, which are mainly used to package and protect items to prevent damage during transportation. Color-printed cartons are a new type of cartons. The corrugated structure inside them is similar to an arch structure, which can prevent shock and has good mechanical properties. With the rapid development of the modern logistics industry, the packaging industry also needs to keep up with the trend and improve the packaging effect of materials to protect the items inside from damage.

[0003] There are still some problems in the existing color printing carton production process: the color printing carton produced by the traditional process has poor waterproof and moisture resistance, and the mechanical strength, flexibility and tensile resistance are also relatively low, which reduces the service life of the carton. Therefore, an intelligent control method and system for the color box production process is expected. Summary of the invention

[0004] The present application provides a method and system for intelligent control of color box production process, which uses deep learning-based image processing technology to process and analyze the image of the back-printed paper collected by the camera, so as to mine the traditional image feature expression and the deep-level image semantic feature expression, thereby combining the complementary feature information of the two to complete the description and characterization of the surface state of the back-printed paper. In this way, the monitoring and identification of printing defects are realized.

[0005] The present application also provides a method for intelligent control of a color box production process, comprising: preparing raw materials, the raw materials comprising corrugated cardboard and slit base paper; printing on surface paper; detecting printing defects; surface treatment; laminating and mounting; die-cutting and forming; and binding into boxes, wherein the detecting printing defects comprises:

[0006] Acquire the image of the paper behind the printing collected by the camera;

[0007] Performing multi-view feature extraction on the post-printing paper image to obtain a post-printing paper fusion image;

[0008] Extracting implicit image features of the post-printing paper fusion map to obtain a post-printing paper fusion feature map;

[0009] Based on the post-printing paper fusion feature map, determine whether there is a printing defect.

[0010] The present application also provides an intelligent control system for color box production process, including: a raw material preparation module, wherein the raw materials include corrugated cardboard and slitting base paper; a face paper printing module; a printing defect detection module; a surface treatment module; a cladding module; a die-cutting module; and a box binding module, wherein the printing defect detection module includes:

[0011] A rear-side paper image acquisition unit, used for acquiring a rear-side paper image captured by a camera;

[0012] A feature extraction unit, used for performing multi-view feature extraction on the back-printed paper image to obtain a back-printed paper fusion image;

[0013] A latent image feature extraction unit, used for extracting latent image features of the post-printing paper fusion image to obtain a post-printing paper fusion feature image;

[0014] The printing defect determination unit is used to determine whether there is a printing defect based on the paper fusion feature map after printing.

[0015] Compared with the prior art, the color box production process intelligent control method and system provided by the present application utilizes deep learning-based image processing technology to process and analyze the image of the back-printed paper collected by the camera, so as to mine the traditional image feature expression and the deep-level image semantic feature expression, thereby combining the complementary feature information of the two to complete the description and characterization of the surface state of the back-printed paper. In this way, the monitoring and identification of printing defects are realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0017] Figure 1 This is a flow chart of an intelligent control method for a color box production process provided in an embodiment of the present application.

[0018] Figure 2 This is a flow chart of the printing defect detection step in a color box production process intelligent control method provided in an embodiment of the present application.

[0019] Figure 3 This is a schematic diagram of the system architecture of a method for intelligent control of a color box production process provided in an embodiment of the present application.

[0020] Figure 4This is a block diagram of an intelligent control system for a color box production process provided in an embodiment of the present application.

[0021] Figure 5 This is an application scenario diagram of an intelligent control method for a color box production process provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the embodiments of the present application more clear, the embodiments of the present application are further described in detail below in conjunction with the accompanying drawings. Here, the illustrative embodiments of the present application and their descriptions are used to explain the present application, but are not intended to limit the present application.

[0023] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as those commonly understood by those skilled in the art of the present application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application.

[0024] In the embodiments of the present application, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or a connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0025] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0026] Color-printed cartons are a new type of cartons, which use a corrugated design similar to an arched structure inside. This design makes the color-printed cartons have good impact resistance and shock absorption. This structure can effectively absorb external impact and protect the items in the carton from damage. This type of carton has been widely used in the modern logistics industry because it can provide more effective packaging protection and meet the requirements for material packaging effects.

[0027] The emergence of color-printed cartons is closely related to the rapid development of the modern logistics industry. With the rapid growth of e-commerce and global trade, the logistics industry has higher and higher requirements for packaging. The design structure of color-printed cartons makes it an ideal packaging choice, which can protect the contents from damage, improve packaging efficiency and reduce transportation costs.

[0028] There are some problems with the current color-printed carton production process, mainly including poor waterproof and moisture-proof capabilities, relatively low mechanical strength, flexibility and tear resistance, and short service life.

[0029] Among them, color-printed cartons made by traditional techniques often cannot effectively protect the items in the box from moisture when facing a humid environment, which limits its application in some humid environments, which means that the stability and protection of the cartons will be affected under humid conditions. Color-printed cartons produced by traditional techniques may show lower mechanical strength and tear resistance when subjected to external pressure and tension, which may make them vulnerable to damage during transportation and handling, affecting the service life and protective effect of the cartons. Due to the existence of the above problems, color-printed cartons made by traditional techniques often have a shorter service life, which means that the cartons need to be replaced more frequently, increasing packaging costs and environmental burdens.

[0030] In response to these problems, the color printing carton production process needs to be continuously improved and innovated to improve the waterproof and moisture-proof ability, mechanical strength, flexibility and tensile resistance of the carton, thereby extending the service life of the carton and improving its applicability in various environments. This may involve the improvement of materials, the optimization of production processes and the introduction of new technologies to achieve higher quality color printing carton production.

[0031] In one embodiment of the present application, Figure 1 Flow chart of a color box production process intelligent control method provided in the embodiment of the present application. Figure 1 As shown, the intelligent control method of the color box production process according to the embodiment of the present application includes: preparing raw materials, the raw materials including corrugated cardboard and cut base paper; surface paper printing; printing defect detection; surface treatment; covering and mounting; die cutting and forming; and binding into boxes.

[0032] The step of preparing raw materials includes corrugated cardboard and slitting base paper. It is necessary to ensure that the quality and specifications of the raw materials meet the production requirements, and check the flatness of the base paper and the thickness of the corrugated cardboard. In this way, the stability and quality consistency in the subsequent production process are guaranteed. In the step of printing the surface paper, the accuracy and clarity of the printed color are controlled to ensure the quality of the printed pattern and text. In this way, the appearance quality of the color-printed carton is improved, and the attractiveness and recognition of the product are enhanced. In the step of detecting printing defects, an automated detection device is used to conduct a comprehensive inspection of the printed product, and printing defects are discovered and corrected in a timely manner. In this way, the quality of the printed product is ensured, the defective rate is reduced, and the production efficiency is improved.

[0033] In the surface treatment step, the printed paper is surface treated, such as coating a protective film or waterproofing. This improves the water resistance and wear resistance of the paper, and increases the service life and protective performance of the color-printed carton. In the cladding and mounting step, the fit and flatness of the cladding are ensured to avoid wrinkles and bubbles. In this way, the appearance quality and hardness of the color-printed carton are enhanced, and the packaging effect and protective performance are improved. In the die-cutting and molding step, the die-cutting size and shape are precisely controlled to ensure the accuracy of the finished product size. In this way, color-printed cartons that meet the specifications are produced, and production efficiency and product consistency are improved. In the stapling and boxing step, suitable stapling equipment is used to ensure that the box is firm and reliable. In this way, high-quality color-printed cartons are produced, and the packaging effect and transportation safety are improved. By strictly controlling the quality and details of each production step, the overall quality and performance of the color-printed cartons can be improved to meet the needs of the market and customers.

[0034] Among them, the step of printing defect detection is an important link in the color box production process. It should be understood that the printing quality directly affects the appearance and product image of the color box. By performing printing defect detection, defects in the printing process can be discovered in time to ensure that the printing quality meets the expected standards. However, traditional printing defect detection usually relies on manual visual inspection, which is not only time-consuming and labor-intensive, but also prone to missed detection or misjudgment due to human negligence or fatigue. Therefore, an optimized solution is expected.

[0035] In response to the above technical problems, the technical concept of this application is to use deep learning-based image processing technology to process and analyze the image of the back-printed paper collected by the camera, so as to mine the traditional image feature expression and the deep-level image semantic feature expression, and then combine the complementary feature information of the two to complete the description and characterization of the surface state of the back-printed paper. In this way, the monitoring and identification of printing defects can be achieved.

[0036] The deep learning model can learn complex image features and semantic information, so it can effectively detect and identify defects in the paper after printing, such as color difference, ink leakage, ink spots, broken lines, etc. This method is more efficient than traditional manual inspection and can greatly reduce the time cost of defect detection. Through deep learning technology, it is possible to achieve a detailed description and characterization of the surface state of the paper after printing, so that the printing quality can be more accurately evaluated, and the product quality and consistency can be improved. Using deep learning technology to monitor and identify printing defects can realize the automation and intelligence of the production process, reduce human intervention, improve production efficiency and reduce labor costs. Through the analysis of a large amount of paper image data after printing, the deep learning model can continuously learn and optimize, so as to achieve continuous improvement of printing quality problems and provide data support for quality control in the production process. Combined with the deep learning model, it is possible to achieve real-time monitoring and early warning of printing quality, timely discover and deal with potential quality problems, reduce the defective rate, and improve production efficiency.

[0037] Using deep learning-based image processing technology to process and analyze paper after printing can bring more efficient and accurate printing defect monitoring and identification, improve product quality and consistency, realize automated production processes, and provide data support for quality improvement, thus bringing significant benefits to the color printing carton production process.

[0038] Figure 2 This is a flow chart of the printing defect detection step in the color box production process intelligent control method provided in the embodiment of the present application. Figure 2 As shown, the printing defect detection step includes: 110, acquiring a printed back paper image captured by a camera; 120, performing multi-view feature extraction on the printed back paper image to obtain a printed back paper fusion map; 130, extracting implicit image features of the printed back paper fusion map to obtain a printed back paper fusion feature map; 140, determining whether there is a printing defect based on the printed back paper fusion feature map.

[0039] In step 110, the image of the paper behind the printing is acquired by the camera, and the camera is set and calibrated to acquire a clear and accurate image of the paper behind the printing, and the influence of light and environmental factors on image acquisition is considered. By acquiring high-quality images of the paper behind the printing as the basis for subsequent processing and analysis, reliable data support is provided for subsequent steps.

[0040] In step 120, the multi-view feature extraction is performed on the back-printed paper image to obtain a back-printed paper fusion map, and a multi-view feature extraction method is used in combination with deep learning technologies such as convolutional neural networks to extract rich features of the back-printed paper image. By obtaining comprehensive and multi-view back-printed paper feature information, a more comprehensive information basis is provided for the analysis and identification in subsequent steps.

[0041] In step 130, the implicit image features of the back-printed paper fusion map are extracted to obtain a back-printed paper fusion feature map, and the implicit features of the back-printed paper fusion map are extracted using a deep learning model, such as by performing feature extraction and abstraction through a convolutional neural network. Obtaining a more abstract and advanced image feature representation is helpful to more accurately describe the state and features of the back-printed paper.

[0042] In step 140, based on the post-printing paper fusion feature map, it is determined whether there is a printing defect, and an effective deep learning model is established. In combination with the post-printing paper fusion feature map, the printing defect is identified and judged, and the accuracy and robustness of the model are considered. By realizing the automated monitoring and identification of printing defects, the accuracy and efficiency of defect detection are improved, which helps to improve product quality and reduce the defective rate.

[0043] By executing the above steps and using deep learning technology to process and analyze the paper after printing, efficient and accurate printing defect monitoring and identification can be achieved, thereby improving product quality and production efficiency.

[0044] Figure 3 Schematic diagram of the system architecture of a color box production process intelligent control method provided in the embodiment of the present application. Figure 3 As shown, the printing defect detection includes: first, obtaining a printed back paper image captured by a camera; then, performing multi-view feature extraction on the printed back paper image to obtain a printed back paper fusion map; then, extracting implicit image features of the printed back paper fusion map to obtain a printed back paper fusion feature map; finally, determining whether there is a printing defect based on the printed back paper fusion feature map.

[0045] Based on this, in the technical solution of the present application, firstly, the image of the paper behind the printing is acquired by the camera. Here, as a common image acquisition device, the camera can easily acquire the image data of the paper behind the printing to obtain the appearance information of the paper behind the printing in real time. Among them, compared with other methods (such as scanners, etc.) to acquire the image of the paper behind the printing, the camera has the following advantages: 1. Real-time: The camera can acquire images in real time and apply them to the subsequent image processing and analysis process. In this way, printing defects can be discovered in time, and corresponding measures can be taken to adjust and correct them. 2. Efficiency: The camera can acquire a large amount of image data in a short time. This is very important for the color box production line with rapid production, which can ensure that each color box has undergone the process of printing defect detection. 3. Convenience: The installation and adjustment of the camera are relatively simple, and it can be easily integrated into the color box production line. And the position can be flexibly adjusted as needed to adapt to color boxes of different sizes and shapes.

[0046] In one embodiment of the present application, multi-view feature extraction is performed on the back-printed paper image to obtain a back-printed paper fusion image, including: edge information feature extraction is performed on the back-printed paper image to obtain an edge detected back-printed paper image; multi-level filtering feature extraction is performed on the back-printed paper image to obtain a filtered back-printed paper image; local grayscale maximum feature extraction is performed on the back-printed paper image to obtain a local grayscale maximum image; and the edge detected back-printed paper image, the filtered back-printed paper image and the local grayscale maximum image are fused to obtain the back-printed paper fusion image.

[0047] It should be understood that, first, edge detection can help identify edge structures and contour information in the image, help extract edge features of the paper image after printing, and provide auxiliary information for subsequent image analysis and processing. Then, multi-level filtering can help smooth the image and highlight specific features in the image, help extract the filtering features of the paper image after printing, and make the image easier to analyze and process. Next, extracting the local grayscale maximum feature helps capture local obvious features in the image, which can help identify important details and texture information in the image. Finally, fusing the edge detection image, the filtered image, and the local grayscale maximum image can comprehensively utilize the information of different features to obtain a more comprehensive and richer paper feature map after printing, providing more information support for subsequent printing defect monitoring and identification. Through these steps, the feature information of the paper image after printing can be extracted more comprehensively, providing a richer data basis for the subsequent deep learning model, and helping to improve the accuracy of monitoring and identifying printing defects.

[0048] Next, the Canny edge detection algorithm is used to perform edge detection on the paper image after printing to obtain edge information features (edge ​​detection of paper image after printing). Here, in the process of defect detection in the printing process, the edge information of the paper image after printing is a very critical feature. It should be understood that printed products usually have clear boundaries and contours, and printing defects often manifest as abnormal changes in edges. Therefore, in the technical solution of the present application, extracting the edge information of the paper image after printing can help the model capture potential printing defects. Among them, the Canny edge detection algorithm is a gradient-based method with good edge positioning and noise suppression capabilities. It can effectively detect strong edges in the image, and refine and connect weak edges to provide clear and accurate edge lines.

[0049] In a specific embodiment of the present application, edge information feature extraction is performed on the post-printing paper image to obtain the edge-detected post-printing paper image, including: using the Canny edge detection algorithm to perform edge detection on the post-printing paper image to obtain the edge-detected post-printing paper image.

[0050] It should be understood that the Canny edge detection algorithm is based on a gradient method that can effectively locate strong edges in an image and has good noise suppression capabilities, which helps to accurately capture edge information in the paper image after printing without being disturbed by noise. The Canny algorithm can refine and connect weak edges to provide clear and continuous edge lines, which is essential for capturing subtle edge changes in the paper image after printing, especially when detecting printing defects. It can more accurately capture abnormal edge changes. The Canny algorithm can produce clear and accurate edge lines, which is very important for subsequent feature extraction and image analysis. Clear edge lines help to extract more representative edge information features and provide more reliable data support for subsequent defect detection.

[0051] That is, using the Canny edge detection algorithm to perform edge detection on the paper image after printing can provide clear and accurate edge information features, help capture abnormal edge changes that may exist in printing defects, and provide an important data basis for subsequent defect detection and identification.

[0052] At the same time, the post-printing paper image is subjected to multi-level filtering processing to obtain multi-level filtering features (post-filtering paper image), wherein the multi-level filtering processing includes Gaussian filtering, median filtering and morphological filtering. Here, the post-printing paper image is subjected to multi-level filtering processing to remove the noise information existing in the post-printing paper image, thereby reducing the detection of printing defects and the judgment of printing quality caused by such noise information. Specifically, Gaussian filtering can smooth the image and retain edge information, median filtering can remove salt and pepper noise and maintain edge clarity, and morphological filtering can improve the shape and texture information of the image. That is, extracting these multi-level filtering features helps to provide more accurate information for subsequent printing defect detection and analysis.

[0053] In a specific embodiment of the present application, multi-level filtering feature extraction is performed on the post-printing paper image to obtain the post-filtered paper image, including: performing multi-level filtering processing on the post-printing paper image to obtain the post-filtered paper image, wherein the multi-level filtering processing includes Gaussian filtering, median filtering and morphological filtering.

[0054] Gaussian filtering can effectively smooth images, remove noise while retaining the edge information of the image, help make the image smoother, reduce noise interference, and provide a clearer image basis for subsequent feature extraction and analysis. Median filtering can effectively remove salt and pepper noise and speckle noise in images, while maintaining the edge information of the image, help further reduce noise interference in the image, and make the image more suitable for subsequent feature extraction and analysis. Morphological filtering can be used to improve the quality of images, remove small interference and discontinuous parts in the image, while maintaining the overall structure and shape of the image, help further clarify the image, and make subsequent feature extraction more accurate. Multi-level filtering processing (including Gaussian filtering, median filtering and morphological filtering) helps to make the paper image clearer after printing, reduce noise interference, and provide a more reliable image basis for subsequent feature extraction, image analysis and defect detection.

[0055] In the actual application scenarios of the present application, the image acquisition environment and conditions are often diverse, including different factors such as light intensity, angle, background, etc. If the model is too sensitive to the illumination changes and noise information in the back-printed paper image, it may cause adverse effects on the detection and identification of printing defects. Therefore, in the technical solution of the present application, the back-printed paper image is further divided into image blocks to obtain a sequence of back-printed paper image blocks, and the grayscale maximum value of each back-printed paper image block in the sequence of back-printed paper image blocks is calculated respectively to obtain a local grayscale maximum feature (local grayscale maximum image). Here, by dividing the back-printed paper image into a plurality of back-printed paper image blocks and calculating the grayscale maximum value of each back-printed paper image block, the grayscale changes in different local areas can be captured. These local grayscale maximum features can help the model locate and identify the location of printing defects. At the same time, the local grayscale maximum feature has a certain robustness to illumination changes and noise. By calculating the grayscale maximum value of each back-printed paper image block, the sensitivity to illumination and noise can be reduced, thereby better capturing the true characteristics of printing defects.

[0056] In a specific embodiment of the present application, extracting the local grayscale maximum feature of the post-printing paper image to obtain a local grayscale maximum image includes: dividing the post-printing paper image into image blocks to obtain a sequence of post-printing paper image blocks; and calculating the grayscale maximum value of each post-printing paper image block in the sequence of post-printing paper image blocks to obtain the local grayscale maximum image.

[0057] Then, the edge information feature, the multi-level filtering feature and the local grayscale maximum feature are fused to obtain a printed back paper fusion map; and the printed back paper fusion map is passed through a feature extractor based on a convolutional neural network model to obtain a printed back paper fusion feature map. That is, the semantic feature information of the paper image at different angles and different viewing angles is comprehensively utilized so that the fused printed back paper fusion feature map has a more outstanding feature expression capability.

[0058] Furthermore, in a specific embodiment of the present application, implicit image features of the back-printed paper fusion map are extracted to obtain a back-printed paper fusion feature map, including: passing the back-printed paper fusion map through a feature extractor based on a convolutional neural network model to obtain the back-printed paper fusion feature map.

[0059] Among them, the feature extractor based on the convolutional neural network model includes: an input layer, a convolution layer, a pooling layer, an activation layer and an output layer.

[0060] In one embodiment of the present application, based on the rear paper fusion feature map after printing, determining whether there is a printing defect includes: optimizing the feature distribution of the rear paper fusion feature map after printing to obtain an optimized rear paper fusion feature map after printing; passing the optimized rear paper fusion feature map after printing through a classifier to obtain a classification result, and the classification result is used to indicate whether there is a printing defect.

[0061] In the technical solution of the present application, since the printed back paper fusion map is obtained by fusing the edge information features, the multi-level filtering features and the local grayscale maximum features, after the printed back paper fusion map passes through a feature extractor based on a convolutional neural network model, each feature matrix of the printed back paper fusion feature map will respectively express the feature semantics corresponding to the edge information features, the multi-level filtering features and the local grayscale maximum features of the printed back paper image. Therefore, there will be sparsity in the representation of semantic features of the channel dimension distribution between each feature matrix of the printed back paper fusion feature map, thereby affecting the integrity of the channel dimension feature distribution of the printed back paper fusion feature map, causing the printed back paper fusion feature map to have difficulty in converging when the class probability is converged through a classifier.

[0062] Therefore, the applicant of the present application preferably first performs a linear transformation on the back-printed paper fusion feature map so that the width and height of the feature matrix are equal, and then optimizes the channel dimension of the converted back-printed paper fusion feature map, which is expressed as: linearly transform the back-printed paper fusion feature map using the following optimization formula so that the width and height of the feature matrix are equal to obtain the converted back-printed paper fusion feature map; optimize the channel dimension of the converted back-printed paper fusion feature map using the following optimization formula to obtain the optimized back-printed paper fusion feature map; wherein the optimization formula is: in, and They are the first and second channel directions of the converted printed back paper fusion feature map. and The feature matrix of the position, and is the scale adjustment hyperparameter, is the first one along the channel direction of the paper fusion feature map after the optimized printing. The feature matrix of the position, is the first one along the channel direction of the converted printed back paper fusion feature map The transposed matrix of the feature matrix of the position, is the first one along the channel direction of the converted printed back paper fusion feature map The inverse of the characteristic matrix of the position, represents positional addition, It means point multiplication by position. Represents matrix multiplication.

[0063] Here, the coupled distribution direction of the local feature distribution of the printed back paper fusion feature map along the channel is predicted in the high-dimensional feature space through the progressive structured embedding calculation of the feature matrix with channel adjacent distribution of the printed back paper fusion feature map, so as to determine the iteratively generated transfer diagram representation based on channel coupling based on the progressive center of the distribution, and reconstruct the contextual relationship of the printed back paper fusion feature map based on the scene layout of the feature matrix by refining the projection normalization proposal from bottom to top along the channel dimension as a whole, so as to enhance the channel dimension integrity of the feature representation of the printed back paper fusion feature map, so as to improve the convergence effect of the printed back paper fusion feature map when the class probability converges through the classifier, and improve the training efficiency and the accuracy of the classification results.

[0064] Then, the optimized post-printing paper fusion feature map is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether there is a printing defect.

[0065] In a specific embodiment of the present application, the optimized post-printing paper fusion feature map is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether there is a printing defect, including: expanding the optimized post-printing paper fusion feature map into a classification feature vector according to a row vector or a column vector; using multiple fully connected layers of the classifier to fully connect encode the classification feature vector to obtain an encoded classification feature vector; and passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0066] By inputting the fusion feature map of the optimized back paper into the classifier for analysis, it is possible to effectively determine whether the printed product has defects. The classifier can learn and identify different types of printing defects, thereby helping to achieve automated defect detection and improve the efficiency and accuracy of the production line. Using a classifier for defect detection can achieve real-time monitoring and judgment. For large-scale production lines, printing defects can be discovered and handled in a timely manner, reducing the defective rate and improving product quality. Defect detection through a classifier can realize an automated production line, reduce manual intervention, reduce labor costs, and improve production efficiency. The classifier can be adjusted and optimized according to actual conditions, continuously improving the accuracy and stability of defect detection, and adapting to different types of printing products and production environments.

[0067] In summary, an intelligent control method for the color box production process based on the embodiment of the present application is explained, which is used to improve the waterproof and moisture-proof ability, mechanical strength, flexibility and tensile resistance of the carton, thereby extending the service life of the carton and improving its applicability in various environments.

[0068] Figure 4 : is a block diagram of an intelligent control system for a color box production process provided in an embodiment of the present application. Figure 4 As shown, the intelligent control system 200 for the color box production process includes: a raw material preparation module 210, wherein the raw materials include corrugated cardboard and slit base paper; a surface paper printing module 220; a printing defect detection module 230; a surface treatment module 240; a tiling module 250; a die-cutting module 260; and a box binding module 270. The printing defect detection module 230 includes: a rear printing paper image acquisition unit 231, which is used to acquire the rear printing paper image collected by a camera; a feature extraction unit 232, which is used to perform multi-perspective feature extraction on the rear printing paper image to obtain a rear printing paper fusion map; an implicit image feature extraction unit 233, which is used to extract implicit image features of the rear printing paper fusion map to obtain a rear printing paper fusion feature map; and a printing defect determination unit 234, which is used to determine whether there is a printing defect based on the rear printing paper fusion feature map.

[0069] Those skilled in the art will appreciate that the specific operations of each step in the above-mentioned color box production process intelligent control system have been referenced above. Figures 1 to 3 The description of the intelligent control method for the color box production process has been introduced in detail, and therefore, its repeated description will be omitted.

[0070] As described above, the color box production process intelligent control system 200 according to the embodiment of the present application can be implemented in various terminal devices, such as a server for intelligent control of the color box production process. In one example, the color box production process intelligent control system 200 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the color box production process intelligent control system 200 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the color box production process intelligent control system 200 can also be one of the many hardware modules of the terminal device.

[0071] Alternatively, in another example, the color box production process intelligent control system 200 and the terminal device may also be separate devices, and the color box production process intelligent control system 200 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0072] Figure 5 This is an application scenario diagram of an intelligent control method for color box production process provided in the embodiment of this application. Figure 5 As shown, in this application scenario, first, an image of the paper after printing captured by a camera is obtained (for example, Figure 5 Then, the obtained paper image after printing is input to a server (for example, Figure 5 In S) as shown in , the server can process the paper image after printing based on the intelligent control algorithm of the color box production process to determine whether there is a printing defect.

[0073] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for intelligent control of color box production process, comprising: Preparing raw materials, including corrugated cardboard and cut base paper; Paper printing; Printing defect detection; Surface treatment; Covered with framed tiles; Die cutting; The binding box is characterized in that the printing defect detection comprises: Acquire the image of the paper behind the printing collected by the camera; Performing multi-view feature extraction on the post-printing paper image to obtain a post-printing paper fusion image; Extracting implicit image features of the post-printing paper fusion map to obtain a post-printing paper fusion feature map; Determining whether there is a printing defect based on the post-printing paper fusion feature map; Wherein, determining whether there is a printing defect based on the post-printing paper fusion feature map includes: Performing feature distribution optimization on the post-printing paper fusion feature map to obtain an optimized post-printing paper fusion feature map; Passing the optimized post-printing paper fusion feature map through a classifier to obtain a classification result, wherein the classification result is used to indicate whether there is a printing defect; The feature distribution of the back-printed paper fusion feature map is optimized to obtain an optimized back-printed paper fusion feature map, including: performing a linear transformation on the back-printed paper fusion feature map so that the width and height of the feature matrix are equal to obtain a converted back-printed paper fusion feature map; performing channel dimension optimization on the converted back-printed paper fusion feature map according to the following optimization formula to obtain an optimized back-printed paper fusion feature map; wherein the optimization formula is: ;in, and They are the first and second channel directions of the converted printed back paper fusion feature map. and The feature matrix of the position, and is the scale adjustment hyperparameter, is the first one along the channel direction of the paper fusion feature map after the optimized printing. The feature matrix of the position, is the first one along the channel direction of the converted printed back paper fusion feature map The transposed matrix of the feature matrix of the position, is the first one along the channel direction of the converted printed back paper fusion feature map The inverse of the characteristic matrix of the position, represents positional addition, It means point multiplication by position. Represents matrix multiplication.

2. The intelligent control method for color box production process according to claim 1 is characterized in that: Performing multi-view feature extraction on the post-printing paper image to obtain a post-printing paper fusion image, including: Extracting edge information features from the paper image after printing to obtain the paper image after edge detection printing; Performing multi-level filtering feature extraction on the post-printing paper image to obtain a post-filtering paper image; Extracting the local grayscale maximum feature of the paper image after printing to obtain a local grayscale maximum image; The edge detection post-printing paper image, the filtering post-printing paper image and the local grayscale maximum image are fused to obtain the post-printing paper fusion image.

3. The intelligent control method for color box production process according to claim 2 is characterized in that: Extracting edge information features from the paper image after printing to obtain the paper image after edge detection printing includes: The edge detection is performed on the post-printing paper image using a Canny edge detection algorithm to obtain the post-printing paper image with edge detection.

4. The intelligent control method for color box production process according to claim 3 is characterized in that: Performing multi-level filtering feature extraction on the post-printing paper image to obtain a post-filtering paper image, comprising: The printed paper image is subjected to multi-stage filtering to obtain the filtered paper image, wherein the multi-stage filtering includes Gaussian filtering, median filtering and morphological filtering.

5. The intelligent control method for color box production process according to claim 4 is characterized in that: Extracting the local grayscale maximum feature of the paper image after printing to obtain a local grayscale maximum image includes: Dividing the post-printing paper image into image blocks to obtain a sequence of post-printing paper image blocks; The grayscale maximum value of each post-printing paper image block in the sequence of post-printing paper image blocks is calculated respectively to obtain the local grayscale maximum value image.

6. The intelligent control method for color box production process according to claim 5 is characterized in that: Extracting implicit image features of the post-printing paper fusion map to obtain a post-printing paper fusion feature map, including: The printed back paper fusion map is passed through a feature extractor based on a convolutional neural network model to obtain the printed back paper fusion feature map.

7. The intelligent control method for color box production process according to claim 6 is characterized in that: The feature extractor based on the convolutional neural network model includes: an input layer, a convolution layer, a pooling layer, an activation layer and an output layer.

8. The intelligent control method for color box production process according to claim 7 is characterized in that: The optimized post-printing paper fusion feature map is passed through a classifier to obtain a classification result, which is used to indicate whether there is a printing defect, including: Expanding the optimized post-printing paper fusion feature map into a classification feature vector according to a row vector or a column vector; Performing full connection encoding on the classification feature vector using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

9. An intelligent control system for color box production process, comprising: A raw material preparation module, wherein the raw materials include corrugated cardboard and cut base paper; a face paper printing module; Printing defect detection module; Surface treatment module; Cladding tile module; Die cutting and forming module; The box binding module is characterized in that the printing defect detection module comprises: A rear-side paper image acquisition unit, used for acquiring a rear-side paper image captured by a camera; A feature extraction unit, used for performing multi-view feature extraction on the back-printed paper image to obtain a back-printed paper fusion image; A latent image feature extraction unit, used for extracting latent image features of the post-printing paper fusion image to obtain a post-printing paper fusion feature image; A printing defect determination unit, used for determining whether there is a printing defect based on the paper fusion feature map after printing; Wherein, determining whether there is a printing defect based on the post-printing paper fusion feature map includes: Performing feature distribution optimization on the post-printing paper fusion feature map to obtain an optimized post-printing paper fusion feature map; Passing the optimized post-printing paper fusion feature map through a classifier to obtain a classification result, wherein the classification result is used to indicate whether there is a printing defect; The feature distribution of the back-printed paper fusion feature map is optimized to obtain an optimized back-printed paper fusion feature map, including: performing a linear transformation on the back-printed paper fusion feature map so that the width and height of the feature matrix are equal to obtain a converted back-printed paper fusion feature map; performing channel dimension optimization on the converted back-printed paper fusion feature map according to the following optimization formula to obtain an optimized back-printed paper fusion feature map; wherein the optimization formula is: ;in, and They are the first and second channel directions of the converted printed back paper fusion feature map. and The feature matrix of the position, and is the scale adjustment hyperparameter, is the first one along the channel direction of the paper fusion feature map after the optimized printing. The feature matrix of the position, is the first one along the channel direction of the converted printed back paper fusion feature map The transposed matrix of the feature matrix of the position, is the first one along the channel direction of the converted printed back paper fusion feature map The inverse of the characteristic matrix of the position, represents positional addition, It means point multiplication by position. Represents matrix multiplication.

Images