Printed matter quality detection method and device and storage medium

Abstract

The invention relates to a printed matter quality detection method and device and a storage medium, and relates to the technical field of printed matter quality detection.The method comprises the steps that material data of a to-be-detected printed matter is obtained; acquiring image data of the to-be-detected printed matter acquired by image acquisition equipment and environment data acquired by environment acquisition equipment; preprocessing the image data according to the material data and the environment data; visual feature data of the preprocessed image data are obtained, the visual feature data at least comprise a printing definition feature, a color consistency feature and a contrast feature, and content feature data of the preprocessed image data are obtained; the content feature data comprises one or more of printed character data features, pattern data features and bar code data features; and obtaining a quality detection result of the to-be-detected printed matter according to the visual feature data and the content feature data of the image data. The method has the effect of improving the quality detection accuracy and stability of the printed matter.

Description

Technical Field

[0001] The present application relates to the technical field of printed matter quality detection, and in particular to a printed matter quality detection method, device and storage medium. Background Art

[0002] Labels are used to identify items, and usually require printing of corresponding text and patterns through printing equipment. In the printing process, in order to improve product quality, printed product quality inspection is an indispensable part. With the continuous growth of market demand and technological advancement, automated inspection systems have gradually become a key tool for improving printed product quality and production efficiency. These systems can not only improve inspection speed and accuracy, but also reduce human errors and reduce production costs, thus bringing significant economic and social benefits to enterprises.

[0003] Among the existing technologies for quality inspection of printed products, commonly used methods include manual visual inspection, traditional optical inspection, and machine vision inspection. Although manual visual inspection is intuitive and reliable, it is subject to human eye fatigue and subjective judgment errors, and it is difficult to ensure consistency and efficiency. Traditional optical inspection relies on specific light sources and sensors. Although it can be partially automated, it has limitations in the detection of complex patterns and multiple colors. Machine vision inspection uses cameras and image processing technology to achieve high-precision automated inspection and is widely used in various fields of the printing industry. However, the existing machine vision inspection system still faces some challenges in practical applications, especially in improving the detection accuracy and stability. For example, the existing inspection system often has difficulty in accurately identifying small color differences, blurs, and misalignments, resulting in inaccurate detection results. In addition, the robustness and adaptability of the system also need to be improved, especially when facing different materials and environmental changes, the performance of the existing system may be unstable. Summary of the invention

[0004] The purpose of this application is to provide a printed matter quality detection method, device and storage medium for improving the accuracy and stability of printed matter quality detection.

[0005] In the first aspect, the present application provides a printed matter quality detection method using the following technical solution: A method for detecting the quality of printed matter, comprising: Obtain material data of the printed product to be tested; Acquire image data of the printed product to be tested acquired by an image acquisition device and environmental data acquired by an environmental acquisition device; Preprocess the image data according to the material data and the environment data; Acquiring visual feature data of the preprocessed image data, wherein the visual feature data at least includes a printing clarity feature, a color consistency feature, and a contrast feature; Acquiring content feature data of the preprocessed image data, wherein the content feature data includes one or more of printed text data features, pattern data features, and barcode data features; The quality inspection result of the printed product to be inspected is obtained according to the visual feature data and the content feature data of the image data.

[0006] By adopting the above technical solution, this application obtains the material data of the printed product to be tested and the current environmental data to pre-process and calibrate the collected image data, and then performs printing quality analysis, which can effectively reduce the impact of the material and environment on the quality detection of printed products, and combines visual feature data with content feature data to perform printed product quality detection. Compared with the traditional single detection method, this method can more accurately and quickly identify various quality problems of printed products, and improve the reliability and efficiency of detection. In particular, the detection capability in complex backgrounds is significantly improved, effectively avoiding misjudgment and missed detection, and can effectively improve the accuracy and stability of printed product quality detection.

[0007] Furthermore, the acquisition of visual feature data of the preprocessed image data includes clarity feature extraction, specifically including: Identify edge clarity of image data through edge detection algorithms; Compare the image data with the standard image data to obtain a structural similarity index; The edge clarity and structure similarity index are input into the pre-built clarity assessment model to obtain the clarity features of the image data.

[0008] By adopting the above technical solution, the present application can ensure the accuracy of printed product quality detection by extracting the clarity characteristics of image data.

[0009] Furthermore, the acquisition of visual feature data of the preprocessed image data also includes color consistency feature extraction, specifically including: Acquiring color spectrum data of the image data by a color measuring device; Convert color spectrum data to standard color space; Calculate the color difference value between the color spectrum data and the standard color spectrum data using the color difference formula; The color difference values ​​are input into a pre-built color consistency evaluation model to obtain the color consistency features of the image data.

[0010] By adopting the above technical solution, the present application performs print quality detection by extracting the color consistency features of image data, thereby ensuring the accuracy of print quality detection.

[0011] Furthermore, the acquisition of visual feature data of the preprocessed image data also includes contrast feature extraction, specifically including: Obtaining a grayscale histogram of the image data to obtain a grayscale distribution of the image data; The maximum contrast of the image data is obtained according to the grayscale difference between the brightest and darkest pixels in the grayscale distribution of the image data; The contrast ratio of the image data is obtained according to the ratio of the maximum gray value to the minimum gray value in the gray level distribution of the image data; According to the grayscale difference between adjacent regions in the grayscale distribution of the image data, a local contrast enhancement algorithm is used to obtain the enhanced local contrast of the image data; The maximum contrast, contrast ratio and local contrast are input into a pre-built color contrast evaluation model to obtain the contrast characteristics of the image data.

[0012] By adopting the above technical solution, the present application performs print quality detection by extracting contrast features of image data, thereby ensuring the accuracy of print quality detection.

[0013] Furthermore, the obtaining of content feature data of the preprocessed image data specifically includes: Determine whether there is text data, pattern data or barcode data through image recognition algorithm; When text data exists, the integrity, clarity and accuracy of the text data are automatically identified through an optical character recognition algorithm to obtain text data features; When pattern data exists, identifying and verifying the integrity, clarity and accuracy of the pattern to obtain pattern data features; When barcode data exists, the readability of the barcode data is identified by barcode recognition software to obtain the barcode data characteristics.

[0014] By adopting the above technical solution, the present application can calibrate the text, patterns and barcodes of printed materials through image recognition, ensure the accuracy of the printed content, and thus improve the comprehensiveness of printed material quality inspection.

[0015] Furthermore, the step of obtaining content feature data of the preprocessed image data further includes: The image data is compared with the standard template through template matching technology to determine whether the printed content on the image data is located in the correct position, and obtain position feature data.

[0016] By adopting the above technical solution, the present application further improves the comprehensiveness of printed product quality inspection by extracting position feature data to determine whether the position of printed content such as text, patterns and barcodes printed on the printed product is accurate.

[0017] Furthermore, obtaining the quality inspection result of the printed product to be tested according to the visual feature data and the content feature data of the image data includes: Construct an initial quality inspection model and configure model parameters of the initial quality inspection model; Obtaining visual feature samples, content feature samples and corresponding quality inspection result samples of different printed products to be tested in a quality inspection database, wherein the quality inspection result samples include normal quality inspection samples and abnormal quality inspection samples; Constructing visual feature samples, content feature samples and corresponding quality inspection result samples of different printed products to be tested into a quality inspection sample set; The initial quality inspection model is trained according to the quality inspection sample set to obtain a trained quality inspection model; The visual feature data and content feature data of the image data are input into the trained quality detection model to obtain the quality detection result of the image data.

[0018] By adopting the above technical solution, the present application constructs a quality inspection model and predicts the quality inspection results through the quality inspection model, thereby improving the intelligence and accuracy of printed product quality inspection.

[0019] Furthermore, the initial quality detection model is trained according to the quality detection sample set to obtain a trained quality detection model, which specifically includes: The quality inspection sample set is divided into a quality inspection training set and a quality inspection test set, and the number of training rounds is preset; Input the quality inspection training set into the quality inspection initial model and perform training according to the preset number of training rounds; When the initial quality inspection model reaches the preset number of training rounds, a trained quality inspection model is obtained; Input the visual feature samples and content feature samples in the quality inspection test set into the trained quality inspection model to obtain quality inspection result prediction samples; The model parameters of the quality detection model are adjusted according to the quality detection result samples in the test set and the corresponding quality detection result prediction samples, and the trained quality detection model is obtained according to the adjusted model parameters.

[0020] By adopting the above-mentioned technical solution, the present application trains the model through a quality inspection sample set, so that the quality inspection model can use deep learning to obtain the feature relationship between visual feature data, content feature data and quality inspection results, thereby improving the intelligence and accuracy of printed product quality inspection.

[0021] In the second aspect, the printed matter quality detection device provided by the present application adopts the following technical solution: A printed matter quality detection device comprises a processor, wherein the processor is used to execute a printed matter quality detection method according to the first aspect; It also includes an image acquisition device and an environment acquisition device. The image acquisition device is used to acquire image data of the printed product to be tested and upload it to the processor, and the environment acquisition device is used to acquire environmental data and upload it to the processor.

[0022] By adopting the above technical solution, the present application collects the image data of the printed product to be tested through the image acquisition device, collects the environmental data of the current environment through the environmental acquisition data, obtains the material data of the printed product to be tested and the current environmental data, performs pre-processing and calibration on the collected image data, and then performs printing quality analysis, which can effectively reduce the impact of the material and environment on the quality detection of printed products, and combines visual feature data with content feature data to perform printed product quality detection. Compared with the traditional single detection method, this method can more accurately and quickly identify various quality problems of printed products, and improve the reliability and efficiency of detection. In particular, the detection capability in complex backgrounds is significantly improved, effectively avoiding misjudgment and missed detection, and can effectively improve the accuracy and stability of printed product quality detection.

[0023] In a second aspect, a computer-readable storage medium provided by the present application adopts the following technical solution: A computer-readable storage medium stores a computer program that can be loaded by a processor and executes a printed product quality detection method as described in the first aspect.

[0024] In summary, the present application includes at least one of the following beneficial technical effects: 1. This application obtains the material data of the printed product to be tested and the current environmental data to pre-process and calibrate the collected image data, and then performs print quality analysis, which can effectively reduce the impact of the material and environment on the quality detection of printed products, and combines visual feature data with content feature data to perform print quality detection. Compared with the traditional single detection method, this method can more accurately and quickly identify various quality problems of printed products, and improve the reliability and efficiency of detection. In particular, the detection capability under complex backgrounds is significantly improved, which effectively avoids misjudgment and missed detection, and can effectively improve the accuracy and stability of printed product quality detection; 2. This application can calibrate the text, pattern and barcode of the printed matter through image recognition to ensure the accuracy of the printed content. By extracting position feature data, the position of the printed content such as text, pattern and barcode printed on the printed matter is accurate, thereby improving the comprehensiveness of the printed matter quality inspection; 3. This application improves the intelligence and accuracy of printed product quality inspection by constructing a quality inspection model and predicting the quality inspection results through the quality inspection model. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a schematic diagram of a printed matter quality detection method according to an embodiment of the present application; Figure 2 is a specific method flow chart of step S6 of an embodiment of the present application; Figure 3 is a connection diagram of a printed matter quality detection device according to an embodiment of the present application; In the figure, 1. processor; 2. image acquisition device; 3. environment acquisition device. DETAILED DESCRIPTION

[0026] The following is combined with Figure 1 -Attached Figure 3 , further details of this application are given.

[0027] Among the existing technologies for quality inspection of printed products, commonly used methods include manual visual inspection, traditional optical inspection, and machine vision inspection. Although manual visual inspection is intuitive and reliable, it is subject to human eye fatigue and subjective judgment errors, and it is difficult to ensure consistency and efficiency. Traditional optical inspection relies on specific light sources and sensors. Although it can be partially automated, it has limitations in the detection of complex patterns and multiple colors. Machine vision inspection uses cameras and image processing technology to achieve high-precision automated inspection and is widely used in various fields of the printing industry. However, the existing machine vision inspection system still faces some challenges in practical applications, especially in improving the detection accuracy and stability. For example, the existing inspection system often has difficulty in accurately identifying small color differences, blurs, and misalignments, resulting in inaccurate detection results. In addition, the robustness and adaptability of the system also need to be improved, especially when facing different materials and environmental changes, the performance of the existing system may be unstable.

[0028] Therefore, the present application provides a method for detecting the quality of printed matter. Figure 1 ,include: S1. Obtain material data of the printed product to be tested.

[0029] Specifically, since different materials of printed products will affect the printing effect, there are different quality inspection standards for printed products of different materials. In the specific implementation process, the material data of the printed product to be tested can be obtained in a variety of ways, such as manual input, scanning of QR codes, RFID tags, etc. The material data may include paper type, coating information, thickness, etc.

[0030] S2. Acquire image data of the printed product to be tested acquired by an image acquisition device and environmental data acquired by an environmental acquisition device.

[0031] Specifically, since different detection environments will affect the detection effect, it is necessary to consider the current environmental data, mainly the brightness data when detecting and collecting image data. In the specific implementation process, the image acquisition device can be a high-resolution camera, such as a CCD or CMOS camera, for taking high-definition images of printed materials. The environmental acquisition device can be a temperature and humidity sensor, a light intensity sensor, etc., which is used to record the conditions of the detection environment. These data help to eliminate external interference and improve detection accuracy.

[0032] S3. Preprocess the image data according to the material data and the environment data.

[0033] Specifically, considering the influence of different material data and environmental data, the image data is preprocessed by denoising, correction, and enhancement. Denoising can use methods such as median filtering and Gaussian filtering to remove noise in the image. Correction can use methods such as geometric correction and color correction to make the image closer to the real situation. Enhancement can use methods such as histogram equalization and Laplace transform to improve the contrast and details of the image.

[0034] S4. Obtain visual feature data of the preprocessed image data.

[0035] Specifically, the visual feature data includes at least printing clarity features, color consistency features, and contrast features.

[0036] S5. Obtain content feature data of the preprocessed image data.

[0037] Specifically, the content feature data includes one or more of printed text data features, pattern data features, and barcode data features.

[0038] S6. Obtaining a quality inspection result of the printed product to be inspected according to the visual feature data and the content feature data of the image data.

[0039] Specifically, by constructing a quality inspection model and inputting the visual feature data and content feature data of the image data into the quality inspection model, the quality inspection result of the printed product to be inspected can be obtained.

[0040] The implementation principle of the embodiment of the present application is: by obtaining the material data of the printed product to be tested and the current environmental data, the collected image data is pre-processed and calibrated, and then the printing quality analysis is performed, which can effectively reduce the impact of the material and environment on the quality detection of printed products, and the printed product quality detection is performed by combining visual feature data and content feature data. Compared with the traditional single detection method, this method can more accurately and quickly identify various quality problems of printed products, and improve the reliability and efficiency of detection. In particular, the detection capability under complex backgrounds is significantly improved, which effectively avoids misjudgment and missed detection, and can effectively improve the accuracy and stability of printed product quality detection.

[0041] In the embodiment of the present application, step S4 includes clarity feature extraction S41, color consistency feature extraction S42 and contrast feature extraction S43.

[0042] The clarity feature extraction S41 specifically includes: S411 . Identify edge clarity of image data using an edge detection algorithm.

[0043] Specifically, edge detection algorithms such as Canny edge detection, Sobel operator, Laplacian operator, etc. can be used to identify the edges in the image. Edge clarity reflects the clarity of the printed product. The sharper the edge, the higher the clarity.

[0044] S412: Compare the image data with the standard image data to obtain a structural similarity index.

[0045] Specifically, the Structural Similarity Index (SSIM) is used to evaluate the similarity between images. SSIM takes into account brightness, contrast, and structural information. This can be achieved through the SSIM algorithm, which compares the image data of the printed product with the standard image data of standard definition. The closer the SSIM value is to 1, the higher the definition.

[0046] S413: Input the edge clarity and the structural similarity index into a pre-built clarity evaluation model to obtain clarity features of the image data.

[0047] Specifically, by extracting the clarity features of image data to detect the quality of printed matter, the accuracy of the printed matter quality detection can be ensured.

[0048] Color consistency feature extraction S42, specifically including: S421. Acquire color spectrum data of the image data through a color measurement device.

[0049] Specifically, use a color measurement instrument, such as a spectrophotometer, to measure the color on the printed product and obtain the color spectrum data. Make sure to measure under a standard light source to eliminate the impact of ambient light on color perception.

[0050] S422, converting the color spectrum data into a standard color space.

[0051] Specifically, the measured spectral data is converted to a standard color space, such as CIELAB, CIELUV or CMYK, for comparison and analysis. The CIELAB color space is based on the perception of the human eye and can more accurately represent color differences.

[0052] S423: Calculate the color difference value between the color spectrum data and the standard color spectrum data using a color difference formula.

[0053] Specifically, the color difference value (ΔE) between the printed color and the standard color is calculated using a color difference formula, such as CIEDE2000 or CIE76. The smaller the color difference value, the better the color consistency.

[0054] S424: Input the color difference value into a pre-built color consistency evaluation model to obtain a color consistency feature of the image data.

[0055] Specifically, by extracting the color consistency feature of the image data to detect the quality of printed matter, the accuracy of the printed matter quality detection can be ensured.

[0056] Contrast feature extraction S43 specifically includes: S431. Obtain a grayscale histogram of the image data to obtain a grayscale distribution of the image data.

[0057] Specifically, the grayscale histogram of the image data is analyzed. The grayscale histogram shows the number of pixels with different gray levels in the image. The contrast can be evaluated by the width and distribution of the grayscale histogram. The wider the grayscale distribution, the higher the contrast.

[0058] S432. Obtain the maximum contrast ratio of the image data according to the grayscale difference between the brightest and darkest pixels in the grayscale distribution of the image data.

[0059] S433. Obtain a contrast ratio of the image data according to a ratio of a maximum grayscale value to a minimum grayscale value in a grayscale distribution of the image data.

[0060] Specifically, contrast indicators such as Max Contrast, Mean Contrast or Contrast Ratio are used to quantify the contrast of an image. The maximum contrast is the grayscale difference between the brightest and darkest pixels in the image, and the contrast ratio is the ratio of the maximum grayscale value to the minimum grayscale value.

[0061] S434. According to the grayscale difference between adjacent regions in the grayscale distribution of the image data, a local contrast enhancement algorithm is used to obtain enhanced local contrast of the image data.

[0062] Specifically, the local contrast of the image is analyzed, that is, the grayscale difference between adjacent regions in the image. The local contrast of the image can be improved by using a local contrast enhancement algorithm, such as color adaptive histogram equalization (CLAHE).

[0063] S435 , input the maximum contrast, contrast ratio and local contrast into a pre-built color contrast evaluation model to obtain contrast features of the image data.

[0064] Specifically, by extracting contrast features of image data to detect printed matter quality, the accuracy of printed matter quality detection can be ensured.

[0065] In another embodiment, the contrast feature can also use a structural contrast evaluation method, such as the contrast part of the structural similarity index (SSIM), to evaluate the contrast of the image structure. The structural contrast takes into account the contrast relationship between different regions in the image, rather than just the difference in grayscale.

[0066] Step S5 of the embodiment of the present application specifically includes: S51. Determine whether there is text data, pattern data or barcode data through an image recognition algorithm.

[0067] S52: When text data exists, the integrity, clarity and accuracy of the text data are automatically identified by an optical character recognition algorithm to obtain text data features.

[0068] Specifically, the text in the image is analyzed to ensure that the text is legible and free of blur, offset, missing or misprinted. Optical character recognition (OCR) technology can be used to automatically identify and verify the correctness of the text.

[0069] S53: When pattern data exists, identify and verify the integrity, clarity and accuracy of the pattern to obtain pattern data features.

[0070] Specifically, identify and verify the integrity of the pattern, without blur, offset, missing or misprinted problems, to ensure the clarity of the pattern.

[0071] S54: When barcode data exists, the readability of the barcode data is identified by barcode recognition software to obtain barcode data features.

[0072] Specifically, use barcode recognition software to check the readability and accuracy of the barcode.

[0073] In order to improve the comprehensiveness of printed product quality detection, step S5 of the embodiment of the present application also includes comparing the image data with the standard template through template matching technology, judging whether the printed content on the image data is located in the correct position, and obtaining position feature data. Check whether the text, pattern, barcode and other elements on the printed product are located in the correct position through position feature data. In the specific implementation process, template matching technology can be used to compare the actual image with the standard template to ensure accurate position. Template matching technology can also be used to check all necessary information on the label, such as product name, net content, production date, shelf life, manufacturer, etc. Ensure that all information is complete and meets relevant laws, regulations and standards.

[0074] like Figure 2 As shown, step S6 of the embodiment of the present application specifically includes: S61. Construct an initial quality inspection model and configure model parameters of the initial quality inspection model.

[0075] Specifically, an optimization scheme detection initial model is constructed based on LSTM, and model parameters of the optimization scheme detection initial model are configured. The model parameters include but are not limited to the number of LSTM units, activation function, and input sequence length.

[0076] S62: Obtain visual feature samples, content feature samples and corresponding quality inspection result samples of different printed products to be tested in the quality inspection database.

[0077] Specifically, the quality inspection database includes quality inspection data stored in previous quality inspection processes, and the quality inspection result samples include normal quality inspection samples and abnormal quality inspection samples.

[0078] S63: constructing a quality inspection sample set by combining visual feature samples, content feature samples and corresponding quality inspection result samples of different printed products to be tested.

[0079] Specifically, visual feature samples and content feature samples are used as input features, and quality detection result samples are used as output features to construct a quality detection sample set to train and test the constructed initial quality detection model.

[0080] S64. Train the initial quality detection model according to the quality detection sample set to obtain a trained quality detection model.

[0081] Specifically, the quality inspection sample set is used to train the initial quality inspection model, in which visual feature samples and content feature samples are used as input data, and quality inspection result samples are used as output data, so that the model can learn the feature relationship between the visual feature samples, content feature samples and the corresponding quality inspection result samples.

[0082] S65. Input the visual feature data and content feature data of the image data into the trained quality detection model to obtain a quality detection result of the image data.

[0083] Specifically, the quality detection model predicts the corresponding quality detection result according to the input visual feature data and content feature data based on the learned feature relationship, and outputs the quality detection result.

[0084] Step S64 of the present embodiment specifically includes: S641. Divide the quality inspection sample set into a quality inspection training set and a quality inspection test set, and preset the number of training rounds.

[0085] Specifically, 70% of the data in the sample set is usually used as a training set, and 30% of the data is used as a test set. The number of training rounds is determined according to actual needs.

[0086] S642: Input the quality inspection training set into the quality inspection initial model, and perform training according to a preset number of training rounds.

[0087] Specifically, the visual feature samples and content feature samples of the quality detection training set are used as input features, and the quantity detection result samples are used as output features, and are all input into the quality detection initial model for training.

[0088] S643: When the initial quality detection model reaches a preset number of training rounds, a trained quality detection model is obtained.

[0089] Specifically, the quality detection initial model autonomously learns the feature relationship between visual feature samples, content feature samples and corresponding quality detection result samples during the training process.

[0090] S644: Input the visual feature samples and content feature samples in the quality detection test set into the trained quality detection model to obtain quality detection result prediction samples.

[0091] Specifically, the quality inspection result prediction samples are predicted by the quality inspection model after preliminary training.

[0092] S645. Adjust the model parameters of the quality detection model according to the quality detection result samples in the test set and the corresponding quality detection result prediction samples, and obtain the trained quality detection model according to the adjusted model parameters.

[0093] Specifically, by comparing the quality inspection result prediction samples and quality inspection result samples predicted by the quality inspection model after preliminary training, the accuracy of the current model after preliminary training can be obtained, and the model can be adjusted according to the comparison results to further improve the model accuracy.

[0094] The implementation principle of the embodiment of the present application is: by constructing a quality inspection model and training the model through a quality inspection sample set, the quality inspection model can use deep learning to obtain the feature relationship between visual feature data, content feature data and quality inspection results, thereby improving the intelligence and accuracy of printed product quality inspection.

[0095] In another embodiment, in order to adapt to the detection requirements under different environmental conditions, the embodiment of the present application can also add a dynamic adjustment mechanism for environmental data. Specifically including: dynamically adjusting the environmental data acquisition frequency, adjusting the acquisition frequency in real time according to the changes in the detection environment, such as increasing the acquisition frequency in an environment with large light changes to ensure the accuracy of the data. In the preprocessing process, an adaptive filter is introduced to dynamically adjust the filter parameters according to the environmental data to adapt to different detection environments. For example, a stronger denoising filter is used in a high-brightness environment, and a weaker denoising filter is used in a low-brightness environment. In the clarity feature extraction, a dynamic threshold adjustment mechanism is introduced to dynamically adjust the threshold of the edge detection algorithm according to the environmental data to adapt to different lighting conditions. In the color consistency feature extraction, an adaptive color correction algorithm is introduced to dynamically adjust the color conversion parameters according to the environmental data to eliminate the influence of ambient light on the color. In the contrast feature extraction, an adaptive contrast enhancement algorithm is introduced to dynamically adjust the local contrast enhancement parameters according to the environmental data to adapt to different contrast requirements.

[0096] By introducing the dynamic adjustment mechanism and adaptive algorithm of environmental data, the embodiments of the present application can better adapt to the detection requirements under different environmental conditions, and improve the flexibility and reliability of detection. Especially in complex production environments, this dynamic adjustment mechanism can effectively reduce external interference and ensure the accuracy of the detection results. In addition, the application of the adaptive algorithm enables the detection system to automatically adjust parameters according to actual conditions, simplifies the maintenance and management of the system, and improves work efficiency.

[0097] In another embodiment, in order to improve the comprehensive performance of detection, the embodiment of the present application can also introduce multimodal data fusion technology to comprehensively utilize multiple types of detection data. Specifically, when obtaining the material data of the printed matter, in addition to the basic material data, the physical property data of the printed matter can also be obtained, such as surface roughness, thickness uniformity, etc., which can be collected by tactile sensors, laser rangefinders and other devices. When acquiring image data and environmental data, in addition to image data and environmental data, sound data, temperature data, etc. can also be obtained, which can be collected by microphones, thermal imagers and other devices. In the preprocessing process, a multimodal data fusion algorithm is introduced to fuse different types of detection data to generate comprehensive feature data. For example, image data and temperature data are fused to generate thermal image data for detecting the temperature distribution of printed matter. In the clarity feature extraction, a multimodal feature extraction algorithm is introduced to comprehensively utilize image data and other types of data to extract richer feature information. For example, the vibration characteristics of the printed matter are extracted in combination with image data and sound data to detect the stability of the printed matter. In the feature extraction of text data, a multimodal feature extraction algorithm is introduced to comprehensively utilize image data and other types of data to extract more accurate text features. For example, the sound features of printed products are extracted by combining image data and sound data to detect the integrity of printed products. In the training process of the quality detection model, a multimodal data fusion strategy is introduced to take different types of detection data as input to generate a more comprehensive feature vector. For example, a multimodal quality detection model is trained by taking image data, temperature data, sound data, etc. as input. By introducing multimodal data fusion technology, the embodiment of the present application can make full use of various types of detection data to improve the comprehensive performance of detection. Compared with the detection method of a single data source, multimodal data fusion can provide richer and more comprehensive information, which helps to more accurately identify the quality problems of printed products. Especially in complex and changeable production environments, multimodal data fusion technology can effectively improve the reliability and accuracy of detection, reduce the misjudgment rate, and improve production efficiency.

[0098] In another embodiment, in order to realize the remote management and maintenance of the printed product quality inspection system, the embodiment of the present application can also introduce remote monitoring and fault diagnosis functions. Specifically, it includes: receiving material data from users through the cloud platform, supporting users to upload material data through mobile phone APP, web pages, etc. Through wireless transmission technology, image data and environmental data are transmitted to the cloud server in real time to support remote monitoring and data analysis. Image data is preprocessed on the cloud server, and powerful computing resources are used to improve the speed and accuracy of preprocessing. Through the cloud computing platform, distributed computing technology is used to accelerate the extraction process of visual feature data and improve detection efficiency. Through the cloud computing platform, distributed computing technology is used to accelerate the extraction process of content feature data and improve detection efficiency. Through the cloud computing platform, distributed computing technology is used to accelerate the training and reasoning process of the quality inspection model and improve detection accuracy. Support users to view the inspection results in real time through mobile phone APP, web pages, etc., and support historical data query and statistical analysis. Through the cloud platform, the operating status of the inspection system is monitored in real time, and remote fault diagnosis and maintenance guidance are supported. Provide fault warning function, when the inspection system is abnormal, notify the user in time and provide solutions. By introducing remote monitoring and fault diagnosis functions, the embodiment of the present application can realize remote management and maintenance of the print quality detection system, improving the availability and reliability of the system. Especially on large-scale production lines, remote monitoring and fault diagnosis functions can timely discover and solve problems, reduce downtime, and improve production efficiency. In addition, through the support of cloud computing platforms, this embodiment can make full use of powerful computing resources, improve the speed and accuracy of detection, and meet the needs of large-scale production.

[0099] In another embodiment, in order to achieve seamless connection between printed product quality inspection and production, the embodiment of the present application can also introduce automated production line integration technology, specifically including: automatically acquiring material data of printed products through sensors on the production line without manual intervention, thereby improving the efficiency of data acquisition. Automatically collecting image data and environmental data through image acquisition equipment and environmental acquisition equipment on the production line, supporting high-speed continuous acquisition. Preprocessing of image data is performed in the embedded system on the production line, and high-performance hardware accelerators are used to improve the speed and accuracy of preprocessing. Through the embedded system on the production line, high-performance hardware accelerators are used to accelerate the extraction process of visual feature data and improve detection efficiency. Through the embedded system on the production line, high-performance hardware accelerators are used to accelerate the extraction process of content feature data and improve detection efficiency. Through the embedded system on the production line, high-performance hardware accelerators are used to accelerate the training and reasoning process of the quality inspection model and improve detection accuracy. Feedback the inspection results to the production line control system in real time to support automated decision-making and control, such as automatic rejection of defective products, automatic adjustment of the production line, etc. Through standardized interfaces and protocols, seamless connection between the quality inspection system and the production line is achieved, supporting plug-and-play. The modular design of the production line is supported, and it can be flexibly configured and expanded according to actual needs to improve the adaptability and flexibility of the system. By introducing the automated production line integration technology, the embodiment of the present application can achieve seamless connection between printed product quality inspection and production, and improve the efficiency and accuracy of inspection. Especially on large-scale production lines, the automated production line integration technology can realize the automated management of the entire process, reduce manual intervention, and improve production efficiency. In addition, through the support of standardized interfaces and protocols, this embodiment can be easily integrated with other production equipment and systems to form a complete production chain and enhance the intelligence level of the entire production system.

[0100] The present application also provides a printed product quality detection device, referring to Figure 1 , including a processor 1, the processor 1 is used to execute a printed product quality detection method: S1. Obtain material data of the printed product to be tested.

[0101] S2. Acquire image data of the printed product to be tested acquired by an image acquisition device and environmental data acquired by an environmental acquisition device.

[0102] S3. Preprocess the image data according to the material data and the environment data.

[0103] S4. Obtain visual feature data of the preprocessed image data.

[0104] S5, obtaining content feature data of the pre-processed image data, S6. Obtaining a quality inspection result of the printed product to be inspected according to the visual feature data and the content feature data of the image data.

[0105] The printed product quality detection device of the embodiment of the present application further includes an image acquisition device 2 and an environment acquisition device 3. The image acquisition device 2 is used to acquire image data of the printed product to be tested and upload it to the processor 1, and the environment acquisition device 3 is used to acquire environmental data and upload it to the processor 1. In the specific implementation process, the image acquisition device 2 can be a high-resolution camera, such as a CCD or CMOS camera, for taking high-definition images of printed products. The environment acquisition device 3 can be a sensor such as a temperature and humidity sensor, a light intensity sensor, etc., for recording the conditions of the detection environment. These data help to eliminate external interference and improve detection accuracy.

[0106] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by the processor 1, the following steps are implemented: S1. Obtain material data of the printed product to be tested.

[0107] S2. Acquire image data of the printed product to be tested acquired by an image acquisition device and environmental data acquired by an environmental acquisition device.

[0108] S3. Preprocess the image data according to the material data and the environment data.

[0109] S4. Obtain visual feature data of the preprocessed image data.

[0110] S5, obtaining content feature data of the pre-processed image data, S6. Obtaining a quality inspection result of the printed product to be inspected according to the visual feature data and the content feature data of the image data.

[0111] When the processor executes the computer program, it can also execute the steps of the printed product quality detection method in any of the above embodiments.

[0112] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM).

[0113] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0114] The embodiments of this specific implementation are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. The same components are represented by the same figure marks. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for detecting the quality of printed matter, characterized in that: include: Obtain material data of the printed product to be tested; Acquire image data of the printed product to be tested acquired by an image acquisition device and environmental data acquired by an environmental acquisition device; Preprocess the image data according to the material data and the environment data; Acquiring visual feature data of the preprocessed image data, wherein the visual feature data at least includes a printing clarity feature, a color consistency feature, and a contrast feature; Acquiring content feature data of the preprocessed image data, wherein the content feature data includes one or more of printed text data features, pattern data features, and barcode data features; The quality inspection result of the printed product to be inspected is obtained according to the visual feature data and the content feature data of the image data.

2. A printed matter quality detection method according to claim 1, characterized in that: The step of obtaining visual feature data of the preprocessed image data includes extracting clarity features, and specifically includes: Identify edge clarity of image data through edge detection algorithms; Compare the image data with the standard image data to obtain a structural similarity index; The edge clarity and structure similarity index are input into the pre-built clarity assessment model to obtain the clarity features of the image data.

3. A printed matter quality detection method according to claim 2, characterized in that: The acquisition of visual feature data of the preprocessed image data also includes color consistency feature extraction, specifically including: Acquiring color spectrum data of the image data by a color measuring device; Convert color spectrum data to standard color space; Calculate the color difference value between the color spectrum data and the standard color spectrum data using the color difference formula; The color difference values ​​are input into a pre-built color consistency evaluation model to obtain the color consistency features of the image data.

4. A printed matter quality detection method according to claim 2, characterized in that: The acquisition of visual feature data of the preprocessed image data also includes contrast feature extraction, specifically including: Obtaining a grayscale histogram of the image data to obtain a grayscale distribution of the image data; The maximum contrast of the image data is obtained according to the grayscale difference between the brightest and darkest pixels in the grayscale distribution of the image data; The contrast ratio of the image data is obtained according to the ratio of the maximum gray value to the minimum gray value in the gray level distribution of the image data; According to the grayscale difference between adjacent regions in the grayscale distribution of the image data, a local contrast enhancement algorithm is used to obtain the enhanced local contrast of the image data; The maximum contrast, contrast ratio and local contrast are input into a pre-built color contrast evaluation model to obtain the contrast characteristics of the image data.

5. A printed matter quality detection method according to claim 1, characterized in that: The obtaining of content feature data of the preprocessed image data specifically includes: Determine whether there is text data, pattern data or barcode data through image recognition algorithm; When text data exists, the integrity, clarity and accuracy of the text data are automatically identified through an optical character recognition algorithm to obtain text data features; When pattern data exists, identifying and verifying the integrity, clarity and accuracy of the pattern to obtain pattern data features; When barcode data exists, the readability of the barcode data is identified by barcode recognition software to obtain the barcode data characteristics.

6. A printed matter quality detection method according to claim 5, characterized in that: The step of obtaining content feature data of the preprocessed image data further includes: The image data is compared with the standard template through template matching technology to determine whether the printed content on the image data is located in the correct position, and obtain position feature data.

7. A printed matter quality detection method according to any one of claims 1 to 6, characterized in that: The method of obtaining the quality inspection result of the printed product to be inspected according to the visual feature data and the content feature data of the image data includes: Construct an initial quality inspection model and configure model parameters of the initial quality inspection model; Obtaining visual feature samples, content feature samples and corresponding quality inspection result samples of different printed products to be tested in a quality inspection database, wherein the quality inspection result samples include normal quality inspection samples and abnormal quality inspection samples; Constructing visual feature samples, content feature samples and corresponding quality inspection result samples of different printed products to be tested into a quality inspection sample set; The initial quality inspection model is trained according to the quality inspection sample set to obtain a trained quality inspection model; The visual feature data and content feature data of the image data are input into the trained quality detection model to obtain the quality detection result of the image data.

8. A printed matter quality detection method according to claim 7, characterized in that: The training of the initial quality detection model according to the quality detection sample set to obtain the trained quality detection model specifically includes: The quality inspection sample set is divided into a quality inspection training set and a quality inspection test set, and the number of training rounds is preset; Input the quality inspection training set into the quality inspection initial model and perform training according to the preset number of training rounds; When the initial quality inspection model reaches the preset number of training rounds, a trained quality inspection model is obtained; Input the visual feature samples and content feature samples in the quality inspection test set into the trained quality inspection model to obtain quality inspection result prediction samples; The model parameters of the quality detection model are adjusted according to the quality detection result samples in the test set and the corresponding quality detection result prediction samples, and the trained quality detection model is obtained according to the adjusted model parameters.

9. A printed product quality inspection device, characterized in that: A method for detecting printed matter quality according to any one of claims 1 to 8 is included; It also includes an image acquisition device and an environment acquisition device. The image acquisition device is used to acquire image data of the printed product to be tested and upload it to the processor, and the environment acquisition device is used to acquire environmental data and upload it to the processor.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and executes a printed product quality detection method as claimed in any one of claims 1 to 8.

Images