Electronic tag production quality control method

Abstract

The invention relates to an electronic tag production quality control method in the technical field of information, and the method comprises the steps: carrying out the noise reduction of an image through a multi-scale denoising algorithm in a second image for vibration interference and dust interference, and obtaining a third image; the online detection system judges the printing quality of the electronic tag according to the OCR recognition result, and if abnormality is detected, an abnormal signal is generated and transmitted to the process control system; the process control system receives the abnormal signal, adjusts operation parameters of printing, fitting and cutting processes, ensures smooth process connection, and avoids unbalance of the production rhythm; comparing the adjusted process parameters with a preset product consistency standard, and if the parameter deviation exceeds a preset range, starting an automatic calibration module to ensure the product consistency; the operation state of the whole production process is monitored in real time through a system monitoring module, if the stability of the system is reduced, an early warning mechanism is triggered, and a maintenance signal is generated and transmitted to a maintenance system.

Description

Technical Field

[0001] The present invention relates to the field of information technology, specifically to the field of label production technology, and particularly to a method for controlling the production quality of electronic labels. Background Art

[0002] In the production process of electronic labels, in order to improve production efficiency and quality and reduce manual intervention, OCR + online detection technology is usually introduced. However, in actual applications, the introduction of these technologies has also brought some new problems and challenges. First, OCR technology has high requirements for image quality. In the production process of electronic labels, due to the complexity of the production environment, such as factors like light changes, vibrations, dust, etc., the quality of the captured images may decline, thereby affecting the accuracy of OCR recognition. Second, online detection technology needs to process a large amount of data in real time, which poses higher requirements for the data transmission and processing capabilities of the system. If the data transmission speed cannot keep up with the production rhythm, or there are delays in data processing, it may affect the normal operation of the production line. In addition, the production of electronic labels involves multiple processes, such as printing, laminating, cutting, etc. The connection and coordination between different processes are also a key issue. If the data between each process cannot be transmitted in a timely and accurate manner, it may lead to a decrease in production efficiency and even quality problems. Finally, the production of electronic labels has high requirements for product consistency and stability. How to ensure the consistency of product quality while guaranteeing production efficiency is also an urgent problem to be solved. Summary of the Invention

[0003] The present invention provides a method for controlling the production quality of electronic labels, including the following steps:

[0004] S101. Obtain a first image during the production process of electronic labels through an image acquisition device, and process the first image using an adaptive illumination compensation algorithm for light change interference to obtain a second image;

[0005] S102. In the second image, for vibration interference and dust interference, perform noise reduction processing on the image using a multi-scale denoising algorithm to obtain a third image;

[0006] S103. Input the third image into a preset OCR recognition model. If there are errors in the OCR recognition result, perform local optimization on the third image using an image enhancement algorithm to obtain a fourth image, and perform OCR recognition again;

[0007] S104. Transmit the OCR recognition result to an online detection system through a high-speed data transmission channel. If the data transmission delay exceeds a preset threshold, start a data compression algorithm to reduce the amount of data and ensure real-time processing capabilities;

[0008] S105. The online detection system determines the printing quality of the electronic tag according to the OCR recognition result. If an abnormality is detected, an abnormal signal is generated and transmitted to the process control system.

[0009] S106. The process control system receives the abnormal signal and adjusts the operating parameters of the printing, laminating, and cutting processes to ensure smooth process connection and avoid imbalance in the production rhythm.

[0010] S107. The operating status data of each process is obtained in real time through the data acquisition module, and the time series analysis method is used to determine whether the production rhythm is stable. If fluctuations are found, the process parameters are dynamically adjusted.

[0011] S108. The adjusted process parameters are compared with the preset product consistency standard. If the parameter deviation exceeds the preset range, the automatic calibration module is activated to ensure product consistency.

[0012] S109. The operating status of the entire production process is monitored in real time through the system monitoring module. If it is found that the system stability decreases, the warning mechanism is triggered, and a maintenance signal is generated and transmitted to the maintenance system.

[0013] The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0014] The present invention discloses a method for controlling the production quality of electronic tags. The method obtains images in the production process through an image acquisition device, and processes the images using an adaptive light compensation and multi-scale denoising algorithm to cope with light changes and vibration interference. The processed images are input into the OCR recognition model, and local optimization is performed if there are errors. The recognition result is transmitted to the online detection system through a high-speed data transmission channel, and the printing quality is judged according to the result. If an abnormality is detected, a signal is generated and transmitted to the process control system to adjust the relevant process parameters. The present invention also obtains the operating status data of each process in real time, judges the stability of the production rhythm through time series analysis, and dynamically adjusts the parameters to ensure product consistency. At the same time, the present invention monitors the entire production process in real time and triggers a warning mechanism when the system stability decreases. Through these measures, the present invention effectively improves the quality control level and production efficiency of electronic tag production. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flowchart of a method for controlling the production quality of an electronic tag according to the present invention.

[0016] Figure 2 is a schematic diagram of a method for controlling the production quality of an electronic tag according to the present invention.

[0017] Figure 3 is another schematic diagram of a method for controlling the production quality of an electronic tag according to the present invention. Detailed implementation manners

[0018] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0019] As Figures 1-3 , a method for controlling the production quality of electronic tags in this embodiment specifically includes the following steps:

[0020] Step S101, obtain a first image during the production process of the electronic tag through an image acquisition device, and process the first image using an adaptive illumination compensation algorithm to obtain a second image in response to light change interference.

[0021] Obtain the original image during the production process of the electronic tag; judge whether the original image is affected by light interference according to a preset light change threshold; if the original image is affected by light interference, trigger the adaptive illumination compensation algorithm, and the adaptive illumination compensation algorithm includes: analyze the brightness distribution characteristics of the original image to determine the illumination compensation parameters; adopt the Retinex algorithm combined with the adaptive histogram equalization method to adaptively adjust the brightness and contrast of the local area of the original image according to the illumination compensation parameters to obtain the compensated image; perform quality evaluation on the compensated image through a convolutional neural network to obtain an evaluation result; dynamically optimize the parameters of the adaptive illumination compensation algorithm according to the evaluation result; transmit the illumination-compensated image to the electronic tag defect detection module.

[0022] Exemplarily, during the production process of electronic tags, light changes can have a significant impact on image quality. To address this issue, an adaptive light compensation algorithm is used to preprocess the original image. First, an electronic tag image on the production line is collected in real time by a high-speed camera and transmitted to the image processing unit. The system pre-sets a light change threshold. For example, if the brightness change exceeds 20%, it is considered to be interfered. When light interference is detected, the adaptive light compensation algorithm is immediately activated. This algorithm calculates the optimal brightness adjustment parameter by analyzing the image brightness histogram. For example, if the overall image is detected to be too dark, the algorithm will appropriately increase the brightness value; otherwise, it will decrease. At the same time, contrast stretching technology is used to enhance image details, making the text and barcodes on the electronic tag clearer and more distinguishable. The Retinex algorithm combined with the adaptive histogram equalization method can effectively handle the problem of uneven illumination. It simulates the human visual system and decomposes the image into an illumination component and a reflection component. By adjusting the illumination component, interference such as shadows and light spots can be eliminated. For example, for an electronic tag image with local overexposure, the algorithm will reduce the brightness of the highlighted area and at the same time enhance the details of the shadow area, thereby obtaining a more balanced visual effect. To evaluate the effect of light compensation, the system uses a convolutional neural network for image quality evaluation. The network model is trained with a large amount of labeled data and can quickly judge image clarity, contrast and other indicators. For example, if the image quality score after compensation is lower than the threshold, the system will automatically adjust the compensation parameters, such as increasing the intensity of contrast enhancement or changing the amplitude of brightness adjustment, to form a closed-loop feedback control. The high-quality image after light compensation processing is then transmitted to the defect detection module. Due to the improvement of image quality, the accuracy of defect detection is significantly improved. For example, for unclear printed text or blurred barcodes, they are more easily recognized in the compensated image. This not only improves production efficiency but also ensures the quality standards of electronic tags. The whole process reflects the important application of computer vision technology in industrial production. Through real-time image processing and adaptive algorithms, the challenges brought by light changes are effectively overcome, providing a reliable guarantee for the quality control of electronic tags. This method is not only applicable to the production of electronic tags but can also be extended to other industrial scenarios that require precise image analysis, such as PCB board detection or packaging printing quality control and other fields.

[0023] Step S102, in the second image, for vibration interference and dust interference, a multi-scale denoising algorithm is used to denoise the image to obtain a third image.

[0024] Obtain a second image to be processed; determine to perform noise reduction processing using a multi-scale denoising algorithm according to the noise type of the second image; obtain image features at different scales by performing multi-scale decomposition on the second image; obtain a preset denoising algorithm for vibration interference and dust interference; adaptively adjust the parameters of the denoising algorithm according to the image features at different scales to obtain an adaptive denoising algorithm at different scales; use the adaptive denoising algorithm at different scales to perform noise reduction processing on the image features at different scales to obtain denoised images at different scales; fuse the denoised images at different scales to obtain a third denoised image; obtain a preset image quality threshold; determine whether the quality of the third image reaches the preset image quality threshold; if so, output the third image as the final noise reduction processing result; if not, return to execute the step of adaptively adjusting the parameters of the denoising algorithm and perform multi-scale noise reduction processing again until the quality of the third image reaches the preset image quality threshold.

[0025] Exemplarily, the multi-scale denoising algorithm is an effective image denoising method. By performing multi-scale decomposition on the image, noise can be analyzed and processed at different scales. Taking wavelet transform as an example, it can decompose the image into low-frequency approximation coefficients and high-frequency detail coefficients. The low-frequency coefficients contain the main structural information of the image, while the high-frequency coefficients contain detail information such as edges and textures as well as noise. In practical applications, a bilateral filter can be used as the basis of the adaptive denoising algorithm. The bilateral filter takes into account both spatial distance and pixel value differences, and can preserve edge information while smoothing the noise. For vibration interference, an adaptive filter based on motion estimation can be designed. This filter first estimates the motion vectors of local regions of the image, and then adjusts the parameters of the filter according to the motion information to better suppress the blurring effect caused by vibration. For dust interference, a method combining morphological operations and median filtering can be adopted. First, the opening operation is used to remove small bright spots (dust particles), and then median filtering is applied to further smooth the image. This combined method can effectively remove dust interference while maintaining the overall structure of the image. During the denoising process, it is crucial to adaptively adjust the algorithm parameters. The local variance can be used as an estimation index for the noise intensity. For regions with strong noise, increase the intensity of the filter; for regions with weak noise, reduce the filtering intensity to retain more details. For example, in bilateral filtering, the standard deviation parameters of spatial and gray values can be dynamically adjusted according to the local variance. Multi-scale image fusion is a key step in the denoising process. The weighted average method can be adopted, and weights are assigned according to the importance of each scale. For example, a higher weight is given to the low-frequency component containing more structural information, while a lower weight is given to the high-frequency component that may contain more noise. This method can maintain the main features of the image while denoising. Image quality assessment is an important link to ensure the denoising effect. The peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM) can be used as evaluation indicators. PSNR reflects the overall quality of the image, while SSIM focuses more on the preservation of structural information. For example, a PSNR greater than 30 dB and an SSIM greater than 0.9 can be set as the quality threshold. If the evaluation result does not meet the standard, it is necessary to return and adjust the denoising parameters, such as increasing the number of iterations or adjusting the filter intensity. Through this multi-scale denoising method, different types of noise interference can be effectively processed, and the image quality can be improved. During the production process of electronic tags, high-quality images are crucial for subsequent defect detection. The denoised image can more accurately reflect the actual condition of the electronic tag, which helps to improve the accuracy and reliability of defect detection, thereby ensuring product quality and production efficiency.

[0026] Step S103: Input the third image into a preset OCR recognition model. If there are errors in the OCR recognition result, locally optimize the third image through an image enhancement algorithm to obtain a fourth image, and perform OCR recognition again.

[0027] Obtain a target image, input the target image into a preset OCR recognition model for text recognition processing to obtain the OCR recognition result of the target image; determine whether there is a recognition error in the OCR recognition result of the target image. If there is a recognition error, determine that the target image needs to be optimized; for the local area with OCR recognition error in the target image, use an image enhancement algorithm to optimize it, and obtain an optimized image through image sharpening and contrast enhancement operations; input the optimized image back into the preset OCR recognition model for text recognition processing to obtain the OCR recognition result of the optimized image; determine whether the recognition accuracy of the OCR recognition result of the optimized image has improved compared to the OCR recognition result of the target image. If the recognition accuracy has improved, determine that the optimization process is effective; extract the text information in the image according to the OCR recognition result of the optimized image, perform structured processing on the text information to obtain structured text recognition result data; store the structured text recognition result data in a specified database.

[0028] Exemplarily, after obtaining the third image, it is input into a preset OCR recognition model for text recognition processing. The OCR model usually adopts deep learning algorithms, such as a combination of convolutional neural network (CNN) and recurrent neural network (RNN), to improve the recognition accuracy. For example, for an image containing handwritten digits, the OCR model may misrecognize "5" as "6". This recognition error may stem from factors such as poor image quality, font deformation, or background interference. For the local areas with recognition errors, an image enhancement algorithm is used for optimization processing. Image sharpening can use the Laplacian operator or Gaussian sharpening filter to enhance the image edges and details. Contrast enhancement can be achieved through histogram equalization or adaptive contrast enhancement (CLAHE). Taking a blurred document image as an example, after applying the sharpening algorithm, the text edges become clearer, which is beneficial for the OCR model to accurately recognize the character shapes. The optimized fourth image is input into the OCR model again for recognition. By comparing the recognition results of the third and fourth images, the effectiveness of the optimization processing is evaluated. For example, for an image with an original recognition rate of 85%, after optimization processing, the recognition rate is increased to 95%, indicating that the optimization processing is indeed effective. This improvement is not only reflected in the overall accuracy but may also be manifested as an improvement in the recognition of specific difficult-to-recognize characters (such as the similar-shaped digits "1" and "7"). Extracting and structuring the text information is to convert the unstructured OCR recognition results into an organized and easily analyzable data format. For example, for the recognized invoice image, information such as commodity name, quantity, unit price, and total price can be extracted and organized into a table form. This structuring processing makes subsequent data analysis and information retrieval more efficient. The structured text recognition results are stored in a specified database to provide data support for subsequent business applications. A relational database (such as MySQL) or a non-relational database (such as MongoDB) can be selected according to the data characteristics and query requirements. For example, for business scenarios that require frequent updates and complex queries, a relational database may be more suitable; while for the storage and rapid reading of a large amount of unstructured data, a non-relational database may have more advantages. This series of processing steps constitutes a complete process for image-based text information recognition and storage. From image acquisition, OCR recognition, image optimization to data structuring and storage, each link is dedicated to improving the recognition accuracy and data availability. This process has a wide range of applications in multiple fields, such as document digitization in automated office systems, bill processing in the financial industry, and parcel information recognition in the logistics industry. By continuously optimizing the algorithms and models, the recognition ability and efficiency of the system can be continuously improved, reducing the labor cost for enterprises and improving the data processing speed and accuracy.

[0029] Step S104: Transmit the OCR recognition result to the online detection system through a high-speed data transmission channel. If the data transmission delay exceeds the preset threshold, start the data compression algorithm to reduce the data volume and ensure real-time processing capabilities.

[0030] Obtain the OCR recognition result, package the OCR recognition result data into a preset format, and send it to the online detection system through a high-speed data transmission channel; the online detection system receives the OCR recognition result data, parses the OCR recognition result data, extracts the OCR recognition result information, and stores the OCR recognition result information in the result cache queue; start a data transmission monitoring thread to obtain the current data transmission delay situation. If the data transmission delay exceeds the preset threshold, trigger the data compression algorithm; the data compression algorithm compresses the OCR recognition result data in the result cache queue, using a lossless compression algorithm to reduce the transmitted data volume while ensuring data integrity; repackage the compressed OCR recognition result data and send it to the online detection system through the high-speed data channel, and update the delay threshold in the data transmission monitoring thread; the online detection system receives the compressed OCR recognition result data, performs decompression processing to restore the original OCR recognition result, and enters the subsequent real-time analysis and processing flow; associate and store the analysis result with the original OCR recognition result, persist it through a distributed cache system, and establish an index of the OCR recognition result and the analysis result based on a time series database.

[0031] Exemplarily, the data transmission and processing of OCR recognition results is a complex process involving multiple technical links. First, the recognition results are packaged into a standard format, such as JSON or XML, to ensure the structuring and readability of the data. A high-speed data transmission channel can use a dedicated optical fiber network or 5G technology to ensure the efficiency of data transmission. After the online detection system receives the data, the parsing process may involve a JSON parser or an XML parser to extract key information such as text content and position coordinates. The result cache queue can be implemented using in-memory databases such as Redis, which supports high-concurrency reading and writing. The data transmission monitoring thread calculates the transmission delay in real time, which can be achieved by calculating the difference between the packet sending time and the receiving time. If the delay exceeds a preset threshold, such as 100 milliseconds, data compression is triggered. The Huffman coding algorithm is selected for compression. Its principle is to allocate different lengths of codes according to the character occurrence frequency. High-frequency characters use short codes, and low-frequency characters use long codes, thereby reducing the overall data volume. The compressed data is repackaged and sent, and at the same time, the delay threshold in the transmission monitoring thread is updated. This dynamic adjustment mechanism can adapt to the network conditions. For example, when the network is congested, the threshold is adjusted to 150 milliseconds, and when the network is smooth, it is adjusted back to 100 milliseconds to balance the transmission efficiency and system load. After the online detection system receives the compressed data, it decompresses the data to restore the original OCR recognition results. This process may involve the reconstruction of the Huffman tree and coding parsing. Subsequently, the system may perform real-time analysis and processing such as text classification and entity recognition. The associated storage of the analysis results and the original OCR recognition results can adopt the form of key-value pairs, where the key is the unique identifier of the OCR result, and the value is the corresponding analysis result. Distributed cache systems such as Memcached can be used for data persistence, providing high-speed reading and writing and fault tolerance capabilities. Time series databases such as InfluxDB are suitable for storing timestamped data, and an index of the OCR recognition time and the result content can be established. This index structure is beneficial for subsequent time range queries and trend analysis, such as counting the change in the text recognition accuracy rate within a certain time period. The design of the entire process aims to balance real-time performance, accuracy, and system resource utilization. Through dynamic compression and distributed storage, the system can maintain stability when the network conditions change, and at the same time facilitate subsequent data mining and analysis. This architecture is not only applicable to OCR systems but can also be extended to other scenarios that require large-scale real-time data processing, such as video stream analysis or Internet of Things data processing.

[0032] Step S105, the online detection system determines the printing quality of the electronic tag according to the OCR recognition result. If an abnormality is detected, an abnormal signal is generated and transmitted to the process control system.

[0033] Obtain the image data of the electronic tag, preprocess the image data, where the preprocessing includes image enhancement and noise removal, to obtain the preprocessed electronic tag image; use OCR technology to perform character recognition on the preprocessed electronic tag image, extract the text content printed on the electronic tag image, and convert the text content into text format data; obtain the preset electronic tag quality assessment rules and abnormal judgment thresholds, and perform quality analysis on the text format data according to the electronic tag quality assessment rules and abnormal judgment thresholds to determine whether there is a printing abnormality in the electronic tag image; if there is a printing abnormality in the electronic tag image, generate abnormal signal data, where the abnormal signal data includes the abnormal type and the detection time; transmit the abnormal signal data to the process control system in real time through a predefined communication protocol and interface; after receiving the abnormal signal data, the process control system takes corresponding measures according to the preset abnormal handling process, and the measures include suspending the production line and isolating the abnormal products.

[0034] Exemplarily, the acquisition of electronic label images is the starting point of quality inspection. High-resolution industrial cameras can capture label details to ensure the quality of basic data for subsequent processing. In the image preprocessing stage, Gaussian filtering is used to remove noise and enhance contrast to improve the clarity of text. For example, for a label image with a resolution of 1920x1080, applying a 3x3 Gaussian kernel for smoothing can effectively reduce image noise while retaining edge information. OCR technology is the core of text recognition. Open-source OCR engines such as Tesseract can efficiently recognize printed text. According to the characteristics of electronic labels, customized models can be trained to improve the recognition accuracy. For example, for labels containing information such as product numbers and production dates, by pre-defining the character set and format template, the recognition accuracy can be increased to over 99%. The design of quality assessment rules is crucial. Scoring criteria can be formulated from multiple dimensions such as character integrity, spacing uniformity, and contrast. Set 85 as the passing line, and an abnormal alarm is triggered when the score is lower than this value. For example, if 1 character is missing or more than 2 characters are blurred in the recognized 12-digit product number, it is determined as a defective product. The abnormal alarm mechanism needs to be fast and accurate. A message queue such as RabbitMQ can be used to achieve real-time transmission of abnormal signals to ensure that the process control system receives feedback at the millisecond level. The signal data should include key information such as the type of abnormality, detection time, and image location for subsequent positioning and processing. After receiving the abnormal signal, the process control system takes corresponding measures according to the preset process. For example, if 3 consecutive defective products are detected, the production line is automatically suspended; for a single abnormality, it can be transferred to the isolation area by a robotic arm. This intelligent processing not only ensures production efficiency but also minimizes the risk of defective products flowing out. Real-time monitoring and quality inspection form a closed-loop management. By establishing a correlation analysis model between production parameters and quality data, potential quality problems can be predicted. For example, when a downward trend in printing quality is detected, the system can adjust the printing pressure or replace the ink in advance to prevent the generation of defective products from the source. This forward-looking control greatly improves production efficiency and product quality. Throughout the process, the real-time and accuracy of data are crucial. High-speed data transmission channels ensure that information is delivered in a timely manner, while distributed storage systems guarantee the reliability and traceability of data. This not only meets the requirements of immediate quality control but also provides valuable data support for subsequent production optimization and quality improvement.

[0035] Step S106, the process control system receives the abnormal signal and adjusts the operating parameters of the printing, laminating, and cutting processes to ensure smooth process connection and avoid imbalance in the production rhythm.

[0036] Obtain the real-time operation status data of each process. When an abnormal signal is received, trigger the automatic adjustment mechanism; according to the preset abnormal handling rules and process parameter ranges, combined with the correlation model established by machine learning algorithms, calculate the optimal adjustment parameters for the printing, laminating, and cutting processes; send the calculated adjustment parameters to the equipment control unit of the corresponding process to modify the operation parameters of the equipment in real time, including printing speed, laminating pressure, and cutting force; use image recognition technology to judge the smoothness of the transfer of semi-finished products between processes after parameter adjustment. If abnormal connection is found, further fine-tune the relevant parameters; dynamically predict the production rhythm according to the real-time production efficiency and material consumption speed of each process, anticipate the situation of rhythm imbalance, and take measures to adjust the production speed of upstream and downstream processes and increase buffer inventory in advance; transmit the production process data to the big data platform in real time, optimize the control strategy and abnormal handling model through data mining and statistical analysis, continuously improve the adaptive ability and control accuracy of the system, and ensure the long-term stable operation of the production line.

[0037] Exemplarily, the process control system is the core of the electronic tag production line, which monitors the operating status of each process in real time to ensure production quality and efficiency. When the system receives an abnormal signal, it will trigger an automatic adjustment mechanism to address various problems in the production process. Taking the printing process as an example, the system may detect an abnormality such as blurred characters. At this time, the control system will calculate the optimal printing parameter adjustment plan according to the preset rules. This may include reducing the printing speed, increasing the ink supply, or adjusting the printing pressure. For example, the printing speed is reduced from 100 sheets per minute to 80 sheets, and the printing pressure is increased by 5% to ensure character clarity. In the lamination process, if it is found that the adhesion strength between the electronic chip and the label substrate is insufficient, the system may increase the lamination pressure and time. The specific adjustment may be to increase the lamination pressure from 2 MPa to 2.5 MPa and extend the lamination time from 1.5 seconds to 2 seconds to improve the adhesion firmness. The adjustment of the cutting process may involve the replacement frequency and cutting force of the cutting tool. For example, when uneven cuts are detected, the system may reduce the cutting speed and increase the cutting force to improve the cutting accuracy. Machine learning algorithms play an important role in parameter optimization. By analyzing historical production data, the system can establish a correlation model between process parameters and product quality. For example, the support vector machine algorithm can help identify the key parameter combinations that cause product defects, while the decision tree algorithm can be used to formulate parameter adjustment strategies under different circumstances. The connection between processes is also a key focus of the system. Through image recognition technology, the system can monitor the transfer status of semi-finished products between processes in real time. If it is found that the transfer is not smooth, such as the label shifting on the conveyor belt, the system will fine-tune the conveyor belt speed or adjust the guiding device to ensure the continuity of the production line. The dynamic prediction of the production rhythm is crucial for maintaining the balance of the production line. The system analyzes the production efficiency and material consumption speed of each process to predict possible bottlenecks. For example, if it is predicted that the speed of the printing process may lead to a shortage of materials in the subsequent lamination process, the system will increase the printing speed in advance or add a buffer inventory between the two processes. Big data analysis plays a key role in continuous optimization. The system transmits various data in the production process, such as equipment operation parameters, product quality indicators, material consumption, etc., to the big data platform in real time. Through data mining technology, the system can discover potential optimization space. For example, the analysis may reveal that a certain specific parameter combination can improve production capacity while ensuring quality, thereby guiding future parameter adjustment strategies. Such an intelligent process control system can not only quickly respond to production abnormalities but also continuously improve the overall efficiency of the production line through continuous learning and optimization, achieving long-term stable and high-quality production of electronic tags.

[0038] Step S107, the operation status data of each process is obtained in real time through the data acquisition module, and the time series analysis method is used to determine whether the production rhythm is stable. If fluctuations are found, the process parameters are adjusted dynamically.

[0039] According to the pre-established data acquisition rules for the operating status of each process, the operating status data of each process on the production line is obtained in real time, and the collected data is stored in the production database in chronological order. For the status data of each process in the production database, a time series analysis algorithm is used to calculate the data characteristic values at each time point, and the calculation results are compared with the preset stable production rhythm threshold to determine whether the current production rhythm is in a stable state. If the time series analysis result indicates that the current production rhythm fluctuates, the dynamic adjustment mechanism of process parameters is triggered. According to the correlation relationship between processes, the target process to be adjusted and its corresponding parameter range are determined. Through the deep reinforcement learning algorithm, combined with historical production data and real-time collected status data, continuously try and optimize the parameter combination of the target process until the optimal parameter value that can restore the production rhythm to stability is found. The optimized process parameter values are sent to the control system of the production line, and by adjusting the corresponding equipment operating parameters, the target process runs according to the optimized parameter values. During the dynamic adjustment of process parameters, continuously monitor the operating status data of each process, and judge whether the adjusted production rhythm has returned to a stable state through the time series anomaly detection algorithm to evaluate the effect of parameter adjustment. If the production rhythm still fails to return to stability within the preset time, the dynamic adjustment process of process parameters is repeated until the production process re-enters a stable state.

[0040] Exemplarily, the acquisition of the operating status data of each process on the production line is the basis for realizing intelligent manufacturing. Taking the printing and packaging production line as an example, parameters such as the ink volume, temperature, and pressure of the printing press, data such as the speed, tension, and temperature of the laminating machine, and information such as the tool position and cutting force of the cutting machine can be collected in real time through a sensor network. These data are stored in the production database at a frequency of seconds or milliseconds to form a continuous time series. The time series analysis algorithm can extract valuable features from these data. For example, calculate the short-term trend of the ink volume of the printing press by the moving average method, or predict the change trend of the tension of the laminating machine by using the exponential smoothing method. Compare these characteristic values with the preset thresholds, such as the ink volume fluctuation of the printing press exceeding ±

[0041] 5%, it can be determined that the production rhythm is abnormal. When an abnormality is detected, the system will trigger the dynamic adjustment mechanism of process parameters. Taking the three processes of printing, laminating, and cutting as an example, if it is found that the unstable printing ink volume leads to a decline in the subsequent lamination quality, the system will first adjust the process parameters of the printing process. The deep reinforcement learning algorithm continuously tries different parameter combinations such as ink supply speed and printing pressure to find the optimal solution that can stabilize the printing quality. During the parameter optimization process, the algorithm will consider the successful experiences in historical production data. For example, in a similar situation, a combination of increasing the ink supply speed by 10% and reducing the printing pressure by 5% has achieved good results, then this set of parameters will be given priority consideration and verification. At the same time, the algorithm will also explore new parameter combinations to adapt to the current specific situation. The optimized parameters will be sent to the production line control system. Taking the printing machine as an example, the control system will correspondingly adjust specific equipment parameters such as the rotation speed of the ink roller and the pressure of the impression cylinder. These adjustments are aimed at making the printing process reach a stable state again, thereby driving the entire production line to resume a stable rhythm. During the parameter adjustment process, the system continuously monitors the data of each process. Through time series anomaly detection algorithms, such as the Local Outlier Factor (LOF) method, it can be judged whether there are still abnormal fluctuations in the adjusted production data. If the production rhythm fails to return to stability within the preset time window (such as 15 minutes), the system will restart the parameter optimization process until a more suitable parameter combination is found. The advantage of this dynamic adjustment mechanism is that it can quickly respond to production fluctuations, reduce manual intervention, and improve production efficiency and product quality stability. Through continuous learning and optimization, the system can continuously improve its ability to handle various production anomalies, providing strong support for the enterprise to achieve intelligent manufacturing and lean production.

[0042] Step S108, compare the adjusted process parameters with the preset product consistency standard. If the parameter deviation exceeds the preset range, start the automatic calibration module to ensure product consistency.

[0043] Obtain the actual parameter values of the current process, compare them with the standard values of the corresponding parameters in the preset product consistency standard, and calculate the parameter deviation value; determine whether the parameter deviation value exceeds the preset allowable range. If it exceeds the allowable range, trigger the automatic calibration module; the automatic calibration module adjusts the corresponding process parameters incrementally or decrementally using the gradient descent algorithm according to the positive or negative situation of the parameter deviation value until the parameter deviation value is reduced to within the preset range; during the parameter adjustment process, continuously obtain the change value of the process parameters, compare it with the preset change threshold, and dynamically control the step size of the parameter adjustment according to the comparison result; after the parameter adjustment is completed, start the trial production program according to the latest process parameter settings and conduct sampling inspection on the product consistency; comprehensively evaluate the quality indicators of the sampled products using the support vector machine algorithm to determine whether the product consistency meets the standard; if the product consistency test result meets the preset standard, save the current process parameter settings as the new standard parameter value; if not, re-execute the parameter automatic calibration process until the product consistency is stabilized at a qualified level; establish a knowledge base for automatic calibration of process parameters, record the optimal parameter combinations and their corresponding product quality data under various process conditions; when the production process or raw materials change, the system automatically retrieves the knowledge base, matches similar working conditions, and calls the corresponding parameter settings to achieve rapid optimization of process parameters.

[0044] Exemplarily, the automatic calibration system of process parameters is the key to ensuring product consistency. The system first obtains the actual parameter values of the current process, such as the temperature, pressure, etc. of the injection molding machine, and compares them with the preset product standards. For example, the standard value of the mold temperature for a certain plastic product is 180°C, with an allowable deviation of ±5°C. The system monitors the temperature in real time. If the detected actual temperature is 188°C, which exceeds the allowable range, the automatic calibration module is triggered. The automatic calibration module uses the gradient descent algorithm to adjust the parameters. In the above example, since the temperature is too high, the system will gradually reduce the heating power. During the adjustment process, the system monitors the temperature change rate in real time. If it is found that the temperature drops too quickly (such as more than 2°C per minute), the adjustment step size is reduced to avoid sudden temperature drops affecting product quality. After the parameter adjustment is completed, the system starts the trial production program. Assuming that plastic bottles are produced, samples will be taken for testing of indicators such as size, weight, and strength. The support vector machine algorithm is used to comprehensively evaluate these indicators to determine whether the product consistency meets the standard. If it is found that key indicators such as the diameter of the bottle mouth and the thickness of the bottle body are within the allowable range, and the difference between batches is less than 1%, the consistency is considered to meet the standard, and the new temperature setting will be saved as the standard parameter value. The establishment of the knowledge base for automatic calibration of process parameters is crucial for coping with changes in production conditions. For example, when the raw material supplier is changed, resulting in a slight change in the melting point of the raw material, the system can quickly retrieve the knowledge base to find the optimal parameter combination in a similar situation. This can not only reduce the debugging time but also ensure the stability of product quality. Through this automated parameter calibration and optimization process, the production efficiency and the consistency of product quality can be greatly improved. It can quickly respond to various changes in the production process, reduce human intervention, and reduce the risk of operation errors. At the same time, by continuously accumulating and updating the knowledge base, the optimization ability of the system will be continuously improved, providing strong support for the long-term stable production of the enterprise.

[0045] Step S109, the system monitoring module monitors the running state of the entire production process in real time. If it is found that the system stability decreases, the warning mechanism is triggered to generate a maintenance signal and transmit it to the maintenance system.

[0046] Acquire equipment operation parameter data collected by sensors deployed at each node of the production process, the equipment operation parameters including temperature, pressure and vibration frequency; use an abnormality detection algorithm to determine whether the equipment operation parameters exceed a preset normal range threshold, and if they exceed the threshold, determine that the equipment operation state is abnormal; if the equipment operation state is abnormal, divide the warning level into at least two levels according to the abnormality degree of the equipment operation parameters, and generate maintenance signals of corresponding levels; for the maintenance signals of different warning levels, obtain corresponding emergency processing strategies from a preconfigured strategy library, and generate equipment control instructions according to the emergency processing strategies; transmit the maintenance signals and related equipment abnormality information to the maintenance system; After receiving the maintenance signal, the maintenance system obtains the equipment location and abnormal parameter information contained in the maintenance signal, and obtains the historical maintenance record of the equipment; uses a case-based reasoning algorithm to generate an equipment maintenance plan according to the equipment location, abnormal parameter information and historical maintenance records; according to the equipment maintenance plan, guides maintenance personnel to perform maintenance work on the equipment, and feeds back the maintenance results to the maintenance system; the maintenance system updates the equipment maintenance record according to the maintenance results, and transmits the maintenance results to the monitoring module; the monitoring module determines whether the equipment has resumed normal operation according to the maintenance results, and if it has resumed normal operation, the warning is lifted, otherwise it continues to monitor until the equipment is fully restored.

[0047] Exemplarily, real-time collection of production equipment operation parameters is the foundation of intelligent manufacturing. Taking an automobile manufacturing plant as an example, a variety of sensors are deployed on the engine assembly line, including temperature sensors to monitor the temperature of the engine cylinder block, pressure sensors to detect oil pressure, and vibration sensors to monitor abnormal vibrations during the assembly process. These sensors transmit data to the central monitoring system in real time through the industrial Internet of Things. The system monitoring module uses statistical-based anomaly detection algorithms, such as the moving average method and the exponential smoothing method, to analyze the equipment parameters in real time. For example, the normal operating temperature range of the engine cylinder block is 80 - 120 °C. If the temperature continuously exceeds 130 °C, the system will determine it as an abnormal state. The design of the early warning mechanism takes into account the degree and duration of the anomaly. Taking the abnormal oil pressure as an example, if the oil pressure drops by 10% for 5 minutes, the system triggers a yellow early warning; if it drops by 20% for 10 minutes, it is upgraded to a red early warning. Different levels of early warnings correspond to different emergency strategies. For example, a yellow early warning may only reduce the production speed, while a red early warning may require shutting down the machine for maintenance. After receiving the early warning signal, the maintenance system will call the case-based reasoning algorithm. This algorithm first retrieves similar cases from historical maintenance records. For example, abnormal engine oil pressure may be related to situations such as oil pump failure and pipeline leakage. The system will match the most similar historical case according to the specific parameters of the current anomaly, such as the amplitude of oil pressure drop and temperature change, and generate corresponding maintenance suggestions. During the equipment maintenance process, maintenance personnel can update the maintenance progress and results in real time through a mobile terminal. For example, if it is found that the abnormal oil pressure is caused by a loose oil pipe joint, after the maintenance personnel tighten the joint, they will enter this information into the system. The system will then update the equipment maintenance record, which not only helps future fault diagnosis but also can be used to optimize the preventive maintenance plan. The entire process embodies the concept of closed-loop control in intelligent manufacturing. From data collection, anomaly detection, early warning triggering to maintenance guidance and result feedback, a complete information flow is formed. This method can not only detect and solve equipment failures in a timely manner but also continuously optimize the production process through data accumulation, improve equipment utilization rate and product quality. For example, by analyzing the frequency of temperature anomalies on the engine assembly line, it may be found that certain batches of parts are more likely to cause overheating problems, thus improving the product design or adjusting the supplier selection at the source. The ultimate goal of this intelligent equipment monitoring and maintenance system is to achieve predictive maintenance. By analyzing a large amount of historical data through machine learning algorithms, the system can predict when the equipment is likely to fail, so as to actively arrange maintenance before the failure occurs, minimize downtime, and improve production efficiency.

[0048] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present application. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for controlling the production quality of electronic labels, characterized in that: The method comprises the following steps: S101, acquiring a first image in the process of producing an electronic label by an image acquisition device, and processing the first image by using an adaptive illumination compensation algorithm to obtain a second image in view of interference from light changes; S102, in the second image, using a multi-scale denoising algorithm to perform denoising on the image in view of vibration interference and dust interference, to obtain a third image; S103, inputting the third image into a preset OCR recognition model, and if there is an error in the OCR recognition result, locally optimizing the third image by using an image enhancement algorithm to obtain a fourth image, and re-performing OCR recognition; S104, transmitting the OCR recognition result to the online detection system through a high-speed data transmission channel. If the data transmission delay exceeds a preset threshold, a data compression algorithm is started to reduce the data volume to ensure real-time processing capability; S105, the online detection system determines the printing quality of the electronic label according to the OCR recognition result, and if an abnormality is detected, an abnormality signal is generated and transmitted to the process control system; S106, the process control system receives abnormal signals and adjusts the operating parameters of the printing, laminating and cutting processes to ensure smooth process connection and avoid imbalance in production rhythm; S107, obtaining the operation status data of each process in real time through the data acquisition module, using the time series analysis method to determine whether the production rhythm is stable, and dynamically adjusting the process parameters if fluctuations are found; S108, comparing the adjusted process parameters with the preset product consistency standards, and if the parameter deviation exceeds the preset range, starting the automatic calibration module to ensure product consistency; S109. The operating status of the entire production process is monitored in real time through the system monitoring module. If the system stability is found to be reduced, the early warning mechanism is triggered, and a maintenance signal is generated and transmitted to the maintenance system.

2. The method for controlling the production quality of electronic labels according to claim 1, characterized in that: The S101 includes: Obtain original images during the production process of electronic labels; Determining whether the original image is disturbed by light according to a preset light change threshold; If the original image is disturbed by light, an adaptive light compensation algorithm is triggered, and the adaptive light compensation algorithm includes: analyzing the brightness distribution characteristics of the original image and determining light compensation parameters; Adopting the Retinex algorithm combined with the adaptive histogram equalization method, according to the illumination compensation parameters, adaptively adjusting the brightness and contrast of the local area of ​​the original image to obtain a compensated image; Performing quality assessment on the compensated image through a convolutional neural network to obtain an assessment result; According to the evaluation result, dynamically optimizing the parameters of the adaptive illumination compensation algorithm; The illumination-compensated image is transmitted to the electronic label defect detection module.

3. The method for controlling the production quality of electronic labels according to claim 1, characterized in that: The S102 includes: Acquire a second image to be processed; Determining, according to the noise type of the second image, to adopt a multi-scale denoising algorithm to perform denoising processing; Obtaining image features at different scales by performing multi-scale decomposition on the second image; Get the preset denoising algorithm for vibration interference and dust interference; Adaptively adjusting the parameters of the denoising algorithm according to the image features at different scales to obtain an adaptive denoising algorithm at different scales; Adopting the adaptive denoising algorithm at different scales to perform denoising on the image features at different scales to obtain denoised images at different scales; Fusing the denoised images at different scales to obtain a denoised third image; Get the preset image quality threshold; Determining whether the quality of the third image reaches the preset image quality threshold; If yes, output the third image as the final denoising result; If not, the process returns to the step of adaptively adjusting the parameters of the denoising algorithm, and performs multi-scale denoising again until the quality of the third image reaches the preset image quality threshold.

4. The method for controlling the production quality of electronic labels according to claim 1, characterized in that: The S103 includes: Acquire a target image, input the target image into a preset OCR recognition model to perform text recognition processing, and obtain an OCR recognition result of the target image; Determine whether there is a recognition error in the OCR recognition result of the target image, and if there is a recognition error, determine that the target image needs to be optimized; For the local area of ​​the target image with OCR recognition error, an image enhancement algorithm is used to optimize the area, and an optimized image is obtained through image sharpening and contrast enhancement operations; Re-inputting the optimized image into the preset OCR recognition model to perform text recognition processing to obtain an OCR recognition result of the optimized image; Determine whether the OCR recognition result of the optimized image has an improved recognition accuracy compared to the OCR recognition result of the target image, and if the recognition accuracy is improved, determine that the optimization process is effective; Extracting text information from the image based on the OCR recognition result of the optimized image, and performing structured processing on the text information to obtain structured text recognition result data; The structured text recognition result data is stored in a designated database.

5. The method for controlling the production quality of electronic labels according to claim 4, characterized in that: The S104 includes: Obtaining OCR recognition results, packaging the OCR recognition result data into a preset format, and sending it to the online detection system through a high-speed data transmission channel; The online detection system receives the OCR recognition result data, parses the OCR recognition result data, extracts the OCR recognition result information, and stores the OCR recognition result information in a result cache queue; Start the data transmission monitoring thread to obtain the current data transmission delay. If the data transmission delay exceeds a preset threshold, trigger the data compression algorithm. The data compression algorithm compresses the OCR recognition result data in the result cache queue, using a lossless compression algorithm to reduce the amount of transmitted data while ensuring data integrity; Repackaging the compressed OCR recognition result data, sending it to the online detection system through the high-speed data channel, and updating the delay threshold in the data transmission monitoring thread; The online detection system receives the compressed OCR recognition result data, performs decompression processing, restores the original OCR recognition result, and enters the subsequent real-time analysis and processing flow; The analysis result is associated with the original OCR recognition result and stored, and is persisted through a distributed cache system, and an index of the OCR recognition result and the analysis result is established based on a time series database.

6. A method for controlling the production quality of electronic labels according to any one of claims 1 to 5, characterized in that: The S105 includes: Acquire image data of the electronic tag, and preprocess the image data, wherein the preprocessing includes image enhancement and noise removal, to obtain a preprocessed electronic tag image; Using OCR technology to perform text recognition on the pre-processed electronic label image, extract the text content printed on the electronic label image, and convert the text content into text format data; Obtaining a preset electronic label quality assessment rule and an abnormality judgment threshold, and performing a quality analysis on the text format data according to the electronic label quality assessment rule and the abnormality judgment threshold to determine whether there is a printing abnormality in the electronic label image; If there is printing abnormality in the electronic label image, abnormal signal data is generated, and the abnormal signal data includes the abnormality type and the detection time; The abnormal signal data is transmitted to the process control system in real time through a predefined communication protocol and interface; After receiving the abnormal signal data, the process control system takes corresponding measures according to a preset abnormality handling process, and the measures include suspending the production line and isolating abnormal products.

7. A method for controlling the production quality of electronic labels according to any one of claims 1 to 5, characterized in that: The S106 includes: Obtain real-time operating status data of each process and trigger the automatic adjustment mechanism when an abnormal signal is received; According to the preset exception handling rules and process parameter range, combined with the association model established by the machine learning algorithm, the optimal adjustment parameters for the printing, laminating and cutting processes are calculated; Send the calculated adjustment parameters to the equipment control unit of the corresponding process to modify the equipment's operating parameters in real time, including printing speed, lamination pressure and cutting force; Image recognition technology is used to determine the smoothness of the transfer of semi-finished products between processes after parameter adjustment. If abnormal connection is found, the relevant parameters are further fine-tuned; According to the real-time production efficiency and material consumption rate of each process, the production rhythm is dynamically predicted to predict the imbalance of the rhythm, and measures such as adjusting the production speed of upstream and downstream processes and increasing buffer inventory are taken in advance; The production process data is transmitted to the big data platform in real time, and the control strategy and exception handling model are optimized through data mining and statistical analysis, so as to continuously improve the system's adaptability and control accuracy and ensure the long-term stable operation of the production line.

8. A method for controlling the production quality of electronic labels according to any one of claims 1 to 5, characterized in that: The S109 includes: Obtain equipment operating parameter data collected by sensors deployed at each node of the production process, wherein the equipment operating parameters include temperature, pressure and vibration frequency; An abnormality detection algorithm is used to determine whether the operating parameters of the device exceed a preset normal range threshold, and if the threshold is exceeded, the operating state of the device is determined to be abnormal; If the operating state of the equipment is abnormal, the warning level is divided into at least two levels according to the abnormal degree of the operating parameters of the equipment, and a maintenance signal of the corresponding level is generated; For the maintenance signals of different warning levels, corresponding emergency processing strategies are obtained from a pre-configured strategy library, and equipment control instructions are generated according to the emergency processing strategies; Transmitting the maintenance signal and related equipment abnormality information to a maintenance system; After receiving the maintenance signal, the maintenance system obtains the equipment location and abnormal parameter information contained in the maintenance signal, and obtains the historical maintenance record of the equipment; Using case-based reasoning algorithms, generating equipment maintenance plans based on the equipment location, abnormal parameter information and historical maintenance records; Instruct maintenance personnel to perform maintenance work on the equipment according to the equipment maintenance plan, and feed back maintenance results to the maintenance system; The maintenance system updates the equipment maintenance record according to the maintenance result and transmits the maintenance result to the monitoring module; The monitoring module determines whether the equipment has resumed normal operation according to the maintenance result. If it has resumed normal operation, the warning is cancelled; otherwise, monitoring is continued until the equipment is fully restored.

Images