Intelligent dough kneading mode selection method and system

Abstract

The invention relates to the technical field of dough kneading control, in particular to an intelligent dough kneading mode selection method and system. The method comprises the following steps: acquiring a real-time dough monitoring image in a dough kneading machine; edge sharpening reconstruction is carried out on the real-time dough monitoring image, sequential frame sequence fitting is carried out, and a dough monitoring sequential frame sequence is constructed; performing frame-by-frame dough contour recognition on the dough monitoring time sequence frame sequence, performing dough time sequence state deep mining, and constructing a dough time sequence state diagram; performing frame-by-frame yeast particle dynamic change mining and multi-moment activity trend prediction on the dough monitoring time sequence frame sequence so as to extract an optimal dough kneading time point; and performing time state matching on the dough time sequence state diagram based on the optimal dough kneading time point, and performing dynamic dough kneading parameter decision to generate dynamic dough kneading parameters. Fine control over the dough kneading machine is achieved, various production environments are automatically adapted, and the intelligent degree of the dough kneading process and the quality of dough are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of dough kneading control, and in particular to a method and system for selecting an intelligent dough kneading mode. Background Art

[0002] With the continuous development of automation technology and the increasing popularity of artificial intelligence, the traditional dough kneading process is gradually developing towards intelligence and automation. In the food production industry such as bread and pastry, the dough kneading process is one of the key steps that affect product quality. The dough kneading process requires that flour and water be fully mixed to form a dough with a certain elasticity and ductility. However, due to differences in flour, temperature, humidity and other factors, the quality and efficiency of dough kneading are easily affected by changes in the external environment. Traditional manual operations or simple mechanical control methods often fail to meet the requirements for accuracy and consistency in the production process.

[0003] During the long kneading process, the kneading effect is closely related to the kneading time, speed, strength and other parameters, and there is great variability and complexity. Traditional kneading control methods usually rely on manual observation and experience judgment, and manually adjust the machine's working mode or stop time based on the operator's experience. This method not only easily leads to inconsistent kneading effects, but also may affect production efficiency and quality stability. In addition, due to the limitations of the mechanical structure and control system of the dough kneading machine itself, traditional kneading control methods often lack real-time monitoring and analysis of the dough state, and it is difficult to effectively adjust the changes in the dough state in the first place. Therefore, a more intelligent kneading mode control selection is needed. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method and system for selecting an intelligent dough kneading mode to solve at least one of the above technical problems.

[0005] To achieve the above object, the present invention provides a method for selecting an intelligent dough kneading mode, comprising the following steps: Step S1: obtaining a real-time dough monitoring image in a dough kneading machine; performing edge sharpening and reconstruction on the real-time dough monitoring image, and performing time series frame sequence fitting to construct a dough monitoring time series frame sequence; Step S2: Recognize the dough contour frame by frame on the dough monitoring time sequence frame sequence, and conduct deep mining of the dough time sequence state to construct a dough time sequence state diagram; Step S3: mining the dynamic changes of yeast particles frame by frame and predicting the activity trend at multiple moments in the dough monitoring time series frame sequence, so as to extract the optimal kneading time point; Step S4: performing time state matching on the dough timing state diagram based on the optimal dough kneading time point, and then making dynamic dough kneading parameter decisions to generate dynamic dough kneading parameters; Step S5: evaluating the dough state based on the dynamic kneading parameters, and fine-tuning and optimizing the instant kneading parameters to construct an instant kneading fine-tuning strategy; Step S6: Perform comprehensive state evaluation based on the instant dough kneading fine-tuning strategy, and conduct self-reinforcement dough kneading performance learning to generate an intelligent dough kneading control model.

[0006] The present invention provides a high-quality, real-time image data source for subsequent analysis through real-time visual monitoring of the dough state. Edge sharpening and reconstruction not only improves the contour recognition accuracy of the dough, but also reduces the problem of image quality affected by ambient light or clutter on the dough surface, thereby enhancing the accuracy of subsequent processing. Through sequential fitting of time series frames, dynamic tracking of dough changes during the kneading process can be achieved, providing a stable time series data basis for subsequent time series state analysis, identifying dough contours frame by frame, ensuring that every detail of the dough during the kneading process can be accurately captured, and providing high-precision data support for analyzing dough morphology changes. Sequence analysis can dig out subtle trends in dough changes, such as expansion, stretching, and rupture during fermentation, help monitor dough status in real time, and build a dough time-series state diagram, which not only improves the understanding of dough state changes, but also provides an important basis for subsequent time dynamic prediction and parameter optimization. Dynamic change mining makes it possible to monitor the activity of yeast in dough in real time, help accurately judge the key changes in the fermentation process, predict the changing trend of yeast activity, and provide a basis for formulating the best kneading time, avoiding excessive kneading or stopping too early, and ensuring that the dough enters the next step in the best state. By identifying the optimal kneading time point, Effectively avoid the phenomenon of over-kneading or under-kneading of dough due to improper time, improve dough quality, based on the precise matching of the optimal kneading time point and dough state, it can perform fine control at different fermentation stages of dough to improve the effect and quality of kneading. This step provides flexible decision support for kneading parameters, and can adjust the kneading intensity and time in real time to ensure that the dough is in the best state. Dynamic kneading parameter decision can better adapt to the actual state of the dough, avoid the potential impact of traditional fixed parameter control methods on dough quality, and improve the consistency and taste of the final product. By real-time monitoring of the dough state, the system can quickly respond to changes in the kneading process to ensure that the dough is always in the best state. Instant fine-tuning can make appropriate adjustments to the real-time state of the dough to avoid excessive or insufficient operation, thereby maximizing the quality of the dough. Through self-reinforcement learning, the system can optimize the control strategy according to the changing dough state, gradually accumulate experience, improve the intelligence level of kneading control, and realize personalized and customized kneading control, so that each batch of dough can get the most suitable kneading parameters, thereby greatly improving the consistency and quality of the product. As time goes by, the system will more and more accurately grasp the relationship between kneading parameters and dough quality, further improving production efficiency and reducing energy consumption.

[0007] Preferably, step S1 comprises the following steps: Step S11: obtaining a real-time dough monitoring image in a dough kneading machine; Step S12: performing edge sharpening and reconstruction on the real-time dough monitoring image to construct a detail-optimized and reconstructed dough image; Step S13: extracting frame-by-frame image changes of the detail-optimized and reconstructed dough image to obtain a dough monitoring image of each frame; Step S14: Calculate the time stamp of sequential frame acquisition for the detail optimized and reconstructed dough image to obtain the acquisition time stamp of each frame; Step S15: performing time-series frame sequence fitting on each frame of the dough monitoring image according to the acquisition timestamp of each frame, and constructing a dough monitoring time-series frame sequence.

[0008] The present invention can obtain sufficiently clear dough state images through high-resolution image acquisition, provide high-quality input data for subsequent detail processing and analysis, and is suitable for various dough kneading machine environments and different dough types, ensuring the wide applicability of the system. The edge sharpening technology can enhance the subtle contours, cracks, bubbles and other details in the dough image, helping to make the subsequent image processing and analysis more accurate. The clear image details enable the subsequent algorithms (such as contour extraction, change detection, etc.) to more accurately identify the shape and state of the dough and reduce recognition errors. Since the shape of the dough may be different during the kneading process, the edge sharpening processing can adapt to images of different dough types and maintain a high recognition effect. Frame-by-frame image change extraction helps to identify subtle changes in each frame of the image, ensuring that the details of the dough state are fully captured and avoiding missing key changes. Through the change extraction of each frame of the image, the dynamic changes of the dough, such as expansion, stretching, rupture, etc., can be monitored in real time, providing detailed tracking of the dough state. Frame-by-frame extraction helps to focus on the dough. State changes, rather than static background, reduce the interference of redundant data and make subsequent analysis more efficient. Timestamp calculation ensures the precise synchronization of image data and time, so that each frame of image corresponds to its corresponding time point one by one, providing accurate time basis for timing analysis and necessary time information for subsequent timing frame sequence fitting, so that the system can perform time-based dynamic analysis and control. By accurately calculating the acquisition time, high-precision time synchronization can be achieved, which helps to avoid analysis deviations caused by time errors. Through timing frame sequence fitting, image data can be converted into time series data, so that each frame of image can be analyzed in the time dimension, enhancing the coherence of subsequent timing analysis. After arranging the images in chronological order, the change process of the dough from the initial state to the finished state can be analyzed, and the key time points and state changes can be identified. The timing frame sequence provides time series data support for subsequent intelligent analysis and decision-making, which can help the system make precise control decisions based on dough state changes.

[0009] Preferably, the specific steps of step S12 are: Perform pixel-by-pixel recognition on the real-time dough monitoring image to extract all pixel points in the image; Perform global pixel average calculation on all pixels in the image to generate the global pixel average of the image; Image noise suppression is performed according to the global pixel average value of the image, thereby obtaining a denoised and optimized dough monitoring image; Perform histogram equalization on the denoised and optimized dough monitoring image to generate an image grayscale histogram; Perform grayscale distribution stretching on the image grayscale histogram and calculate the grayscale value of the stretched pixel; performing contrast enhancement according to the grayscale values ​​of the stretched pixels to obtain a contrast enhanced dough monitoring image; Perform edge gradient information analysis on the texture detail enhanced dough monitoring image to obtain an edge map; Performing slight texture change sharpening on the edge map to generate a slight texture change sharpened edge map; The contrast enhanced dough monitoring image is reconstructed by edge sharpening according to the micro-texture change sharp edge map to construct a detail optimized reconstructed dough image.

[0010] The present invention can ensure that every detail in the dough image is fully captured through pixel-by-pixel processing, providing a basis for subsequent image optimization, feature extraction and analysis. Accurate pixel data makes subsequent denoising, enhancement and other processing more accurate, and can effectively remove redundant data and extract subtle changes in the dough state. Through pixel-level processing, it can adapt to changes in different lighting conditions, dough types or environmental factors, ensuring that the system can maintain high-precision processing effects under various conditions. The global pixel average can reflect the overall brightness level of the image, helping the system to identify whether the image needs to be adjusted or enhanced in brightness. Through global pixel averaging, unimportant details in the image can be effectively filtered out and image noise can be reduced. Impact on subsequent processing: Noise suppression can effectively remove stray signals in the image, improve image quality, and provide clearer images for subsequent detail processing and analysis. After removing noise, the actual shape and state of the dough are clearer, and details are easier to capture, thereby improving the image recognition. The denoising and optimized image can better adapt to subsequent analysis and feature extraction, reduce errors and deviations, and improve the accuracy of overall intelligent control. Histogram equalization enhances the contrast of the image, making the details of the dough more obvious, especially in low contrast or insufficient light conditions. The equalized image can enhance the details of important areas in the image without changing the original image content, which helps with subsequent feature extraction. ,The balancing of contrast makes the different layers of dough (such as surface texture, bubbles, cracks, etc.) easier to identify, improving the visual effect of the image. Through grayscale stretching, the contrast of tiny details in the image is enhanced, making the different states of the dough (such as bubbles, cracks, etc.) more prominent and convenient for analysis. The image after grayscale stretching helps to highlight the features in the image, making subsequent image recognition and analysis more accurate. The stretching process expands the grayscale range, effectively improves the grayscale contrast of the image, and enhances the expressiveness of the dough in the image. Contrast enhancement helps to highlight the key features of the dough, such as surface texture, bubbles, etc., which is convenient for subsequent recognition and analysis. After enhancing the contrast, the details of the dough are clearer in the image. The edge map extracts subtle changes in the dough and enhances the visualization of the dough surface structure and morphology, facilitating subsequent analysis. Sharpening the edge map can enhance subtle changes in the image, making the subtle structure, bubbles, cracks, etc. of the dough more obvious, which is conducive to accurately monitoring the dough state. Sharpening subtle texture changes improves the recognizability of image details and helps to more accurately analyze the dough state and defects. The sharpened edge map can provide more detailed texture information, which helps the system to judge the fermentation, stretching and forming state of the dough.

[0011] Preferably, the specific steps of step S2 are: Step S21: performing dough contour recognition frame by frame on the dough monitoring time series frame sequence, and extracting the dough contour line of each frame; Step S22: Tracking the dynamic changes of the dough contour line of each frame to obtain a dynamic change trajectory of the dough contour; Step S23: performing contour change trend analysis on the dynamic change trajectory of the dough contour to generate dough contour change trend characteristics; Step S24: evaluating the dough surface roughness of the dough monitoring time series frame sequence to generate a dough surface roughness time series curve; Step S25: performing frame-by-frame analysis of the dough crack generation situation on the dough monitoring time series frame sequence to obtain crack evolution characteristics in the dough; Step S26: Conduct in-depth exploration of the dough time-series state based on the dough surface roughness time-series curve, dough contour change trend characteristics, and dough crack evolution characteristics to construct a dough time-series state diagram.

[0012] The present invention extracts the contour line of the dough frame by frame, and the system can accurately obtain the shape and size changes of the dough, and capture the morphological changes of the dough during the kneading process. Contour recognition can help the system better understand the surface characteristics of the dough, detect details such as cracks and bubbles, and provide data support for subsequent analysis. The contour recognition of each frame provides basic data for subsequent dynamic analysis, helping the system to track the change trend of the dough in real time. Dynamic tracking of contour changes can help the system observe the shape changes of the dough in real time, and can detect key processes such as expansion and deformation. Contour change trend analysis can predict the future morphological changes of the dough and identify the change pattern of the dough state, thereby providing an important reference for subsequent control strategies. By analyzing the contour change trend, it can be timely discovered whether the dough has reached the optimal kneading state, so as to adjust the kneading process to achieve the best effect. The dough contour change trend provides an important basis for the intelligent decision-making of the system, and can realize dynamic regulation based on the dough state. Roughness evaluation can reflect the transition of the dough surface from smooth to rough during the kneading process. This change is very important for the fermentation and molding of the dough. The roughness curve can be used as an important indicator for judging the quality of the dough, helping the system to understand whether the dough is in the best state. Based on the roughness curve, the dough surface can be adjusted to achieve the best effect. The system can monitor the surface state of the dough in real time to make more accurate kneading adjustments. The generation and evolution of cracks are crucial to the quality and kneading process of the dough. By tracking the crack changes frame by frame, the system can promptly identify whether the dough has abnormal cracks to avoid over-kneading or stretching of the dough. The crack evolution characteristics help to realize dynamic monitoring of the dough state and avoid dough cracking problems caused by over-kneading. Crack detection can provide data support for parameter adjustment during the kneading process to ensure that the dough shape and structure remain optimal. By combining multiple features (such as roughness, contour changes and crack evolution), the system can accurately identify whether the dough has abnormal cracks and avoid over-kneading or stretching. The dough timing state diagram provides the system with a global view of the dough state, helping the system to comprehensively evaluate the overall state of the dough and make more accurate control decisions. By deeply mining the dough timing state, it can provide a scientific basis for the selection of intelligent kneading mode, ensuring that the kneading strength and time adjustment at each stage can achieve the best effect. The deeply mined dough timing state diagram can help the system discover potential trends in dough state changes and make adjustments in advance, thereby optimizing the kneading process and avoiding over-kneading or under-kneading.

[0013] Preferably, step S3 specifically comprises the following steps: Step S31: mining the dynamic changes of yeast particles frame by frame in the dough monitoring time series frame sequence to extract the dynamic change characteristics of yeast particles; Step S32: performing yeast temporal activity analysis according to the dynamic change characteristics of yeast particles to generate a yeast temporal activity curve; Step S33: Evolving the activity variation law of the yeast time series activity curve to obtain the yeast activity variation law; Step S34: performing multi-time activity trend prediction based on the yeast activity change rule, thereby generating yeast activity trend prediction data at multiple times; Step S35: Calculate the optimal kneading time based on the yeast activity trend prediction data at multiple moments, so as to extract the optimal kneading time point.

[0014] The present invention can accurately track the distribution and quantity changes of yeast and capture the details of yeast activity fluctuations by mining the dynamic changes of yeast particles frame by frame. By identifying the distribution and movement of yeast particles in the dough, the reproduction, consumption and activity change process of yeast can be understood, providing reliable data for subsequent fermentation analysis. Detailed particle dynamic mining can reveal the slight changes in the activity of yeast particles, which helps to adjust the parameters in the dough kneading process in time and avoid premature or late activity imbalance. Through the time-series activity curve, the fluctuation of yeast activity during the dough kneading process can be clearly seen, helping operators or systems to grasp the changes in yeast activity. The curve can identify the time period when the yeast activity reaches a peak, thereby determining the most suitable time for dough kneading and ensuring that the dough fermentation achieves the best effect. By drawing the yeast time-series activity curve in real time, the system can adjust the kneading intensity and time in the dough kneading process in time according to the activity fluctuations to achieve more accurate fermentation control. Through the evolution analysis of the yeast activity curve, the periodic or trend change law of yeast activity can be extracted, providing theoretical support for the subsequent fermentation process. By understanding the changing rules of yeast activity, the system can foresee the key moments of dough fermentation, thereby dynamically adjusting the kneading time to avoid over-kneading or under-kneading. The analysis of the changing rules of yeast activity can further accurately predict the key nodes in the dough fermentation process, thereby providing a more refined control strategy. Through multi-time activity prediction, the system can grasp the changing trend of yeast activity in advance, thereby pre-regulating the dough fermentation process to avoid delayed response. The predicted yeast activity trend can provide the system with a basis for adjusting the kneading strategy in advance, ensuring that the yeast activity fluctuates within an appropriate range to achieve the best fermentation effect. The extraction of the optimal kneading time point can ensure that the yeast activity is in the most appropriate state, avoiding premature or late kneading operations that lead to insufficient or over-fermentation of the dough. By accurately calculating the optimal kneading time, the fermentation effect of the dough can be maximized, time waste can be reduced, and production efficiency can be improved. The optimal kneading time can ensure that the dough reaches the best fermentation state during the kneading process, thereby improving the final quality indicators of the dough such as taste and elasticity.

[0015] Preferably, the specific steps of step S35 are: Perform multi-time activity trend peak calculation on yeast activity trend prediction data at multiple times, and extract the activity trend peak of each time period; The activity trend peak in each time period is used to identify the inflection point of activity change and marked as the highest peak point of yeast activity; Define preset kneading time parameters; Calculate the kneading buffer time at the highest peak point of yeast activity according to the preset kneading time parameters to obtain the kneading buffer time at the highest peak point; The optimal kneading time is calculated based on the kneading buffer time at the highest peak point to obtain the optimal kneading time point.

[0016] The present invention can accurately identify the highest point of yeast activity in each time period through peak value calculation, which is helpful to understand the period when yeast is most active, which is very important for the most appropriate kneading time in the fermentation process. By extracting peak value data, the system can provide a clear decision basis for operators or automation systems, and help make more accurate kneading time selection. The activity peak value of each time period provides the system with a key point of periodic change of yeast activity, which is helpful to optimize key operations in the dough fermentation process. By identifying the inflection point, the peak of yeast activity change can be clearly identified to help determine the best kneading time for the dough. This moment is the moment when the dough has the greatest fermentation potential, so it is very critical. The identification of the inflection point can further accurately select the kneading time to avoid missing the best kneading time, thereby avoiding insufficient or excessive yeast activity consumption. By identifying the highest peak point of yeast activity, the fermentation process can be controlled more accurately to prevent yeast activity from reaching the peak too early, resulting in incomplete fermentation, or reaching the peak too late, resulting in delayed kneading time. The preset kneading time parameter provides a standardized control specification for the kneading process, which is helpful to ensure the consistency and repeatability of the kneading operation. According to different dough types and fermentation requirements, the kneading time can be adjusted. The whole kneading time parameter enables the system to adapt to various production needs, thereby ensuring the optimal fermentation effect of the dough. The preset kneading time can provide a reference value for the system to avoid kneading time that is too short or too long due to improper operation, and ensure the best kneading conditions. By calculating the buffer time, it is possible to avoid kneading the dough immediately after the yeast activity reaches its peak, and prevent excessive consumption of yeast activity due to premature kneading. Reasonable calculation of the buffer time ensures that the yeast activity is fully utilized and provides appropriate time for subsequent kneading, so that the dough can obtain the best fermentation effect. The setting of the kneading buffer time can be The system or operator is provided with greater flexibility to adjust the start time of dough kneading according to actual needs to avoid fluctuations in dough state. By combining the buffer time and the highest peak point, the system can accurately calculate the optimal dough kneading time point to avoid uneven dough fermentation due to kneading operations too early or too late. The optimal dough kneading time point ensures that the yeast activity is in the most appropriate state, so that the dough can obtain the best fermentation effect during the kneading process and improve the quality of the final product. The precise dough kneading time point can not only improve the fermentation effect of the dough, but also improve the efficiency of the production line, reduce time waste and increase production capacity.

[0017] Preferably, the specific steps of step S4 are: Step S41: performing time state matching on the dough time sequence state diagram based on the optimal kneading time point, and extracting the dough time sequence state characteristics at the optimal kneading time point; Step S42: Analyze the optimal dough state of the dough time sequence state diagram to generate optimal dough state characteristics; Step S43: performing an optimal state deviation analysis on the optimal dough state characteristics based on the dough temporal state characteristics at the optimal dough kneading time point to obtain the dough state deviation characteristics at the optimal dough kneading time point; Step S44: making dynamic kneading parameter decisions based on the dough state deviation characteristics at the optimal kneading time point to generate dynamic kneading parameters.

[0018] The present invention can accurately capture the dough characteristics corresponding to the optimal kneading time point through time state matching, and provide basic data for subsequent analysis and adjustment. The time sequence state characteristics of the optimal kneading time point provide necessary data support for dynamically adjusting the kneading parameters, ensuring that the kneading process is carried out in the best state. Determine the dough state characteristics at the optimal time point to avoid operation delays or premature decisions, and ensure that the dough is kneaded in the most suitable fermentation stage. Through a comprehensive analysis of the dough time sequence state diagram, the overall quality of the dough can be scientifically evaluated to ensure that the dough is kneaded in the best fermentation state. The optimal dough state characteristics provide the standard of the ideal state of the dough, and the system can make timely adjustments to the kneading process according to these characteristics, so as to obtain better fermentation effect and final quality. Through data-driven optimal dough state analysis, errors in manual operation can be reduced, ensuring that each operation can be carried out according to accurate dough state characteristics. Through deviation analysis, the gap between the dough state and the expectation can be discovered in time, potential fermentation or kneading problems can be identified, and the dough quality can be prevented from decreasing due to excessive deviation. The dough state deviation characteristics provide an adjustment basis for the system, and the system can automatically or manually adjust the parameters of the kneading process according to these deviations to ensure that the dough finally reaches the ideal state. By analyzing the deviation characteristics, the system can dynamically adjust the kneading strategy under different production environments or operating conditions, so as to flexibly respond to various changes. The generation of dynamic kneading parameters can automatically adjust the various parameters in the kneading process according to the real-time dough state and deviation characteristics to achieve adaptive control. Dynamic adjustment of kneading parameters can effectively optimize the dough kneading process, improve kneading efficiency, and avoid problems such as over-kneading or insufficient fermentation of the dough. This step makes the kneading process not only rely on fixed rules, but also dynamically adjusts according to real-time feedback and data, which improves the intelligence of the production process and reduces manual intervention.

[0019] Preferably, the specific steps of step S5 are: Step S51: performing real-time dough kneading control based on dynamic dough kneading parameters, and acquiring real-time dough kneading monitoring images; Step S52: performing dough shape analysis on the real-time dough kneading monitoring image to obtain the real-time dough kneading shape; Step S53: performing dough state evaluation on the real-time dough shape during kneading to generate a real-time dough state evaluation value; Step S54: Based on the real-time kneading state evaluation value, the dynamic kneading parameters are fine-tuned and optimized to construct an instant kneading fine-tuning strategy.

[0020] By acquiring real-time dough kneading monitoring images, the present invention enables the system to timely capture changes in the shape, texture and state of the dough during the kneading process, thereby providing accurate visual data for subsequent analysis; by performing real-time dough kneading control based on dynamic dough kneading parameters, the system can automatically perform adjustments to ensure that the dough kneading process is carried out according to the optimal strategy without relying on manual intervention, thereby improving the level of automation and accuracy; the real-time monitoring images provide an effective means for continuous tracking of the dough state, helping the system to quickly determine whether it deviates from the target state and take timely adjustment measures; through real-time analysis of the dough shape, the system can determine whether the dough has irregular deformation or rupture during the kneading process, and then analyze the processing effect of the dough, which is the basis for adjusting the dough kneading parameters; if the shape of the dough changes abnormally (such as unevenness, overstretching or collapse), the system can promptly discover the problem and take corresponding adjustment measures to avoid the deviation in the dough kneading process affecting the final dough quality; the shape of the dough is the key to the quality of the dough. An important sign, shape analysis can intuitively reflect the state of the dough, thus providing an effective basis for subsequent control and optimization. By converting the shape changes of the dough into quantitative state evaluation values, the system can accurately evaluate the dough in a data-driven manner, avoiding the uncertainty in traditional empirical operations. Real-time evaluation values ​​can quickly feedback the state of the dough, providing a scientific basis for dynamically adjusting the kneading parameters to ensure that the dough is always in the best state. Through instant fine-tuning of the kneading parameters, the system can optimize the dough state according to real-time feedback, avoid excessive kneading of the dough or uneven fermentation, and ensure the best quality of the final dough. The instant fine-tuning strategy enables the system to flexibly adapt to different dough types, environmental changes and other factors according to the dynamic changes of the dough state, thereby realizing intelligent kneading control. Real-time adjustment of kneading parameters can maximize the kneading effect and production efficiency, avoid affecting product quality due to too long or too short kneading time, thereby saving production time and improving production capacity.

[0021] Preferably, the specific steps of step S6 are: Step S61: extracting a final dough kneading image according to the real-time dough kneading monitoring image; Step S62: performing a comprehensive state evaluation on the final dough kneading image according to the optimal dough state characteristics to obtain a final dough state evaluation value; Step S63: Based on the final dough state evaluation value, the instant dough kneading fine-tuning strategy is self-reinforced to learn the dough kneading performance, thereby generating an intelligent dough kneading control model.

[0022] By extracting the final kneading image, the system of the present invention can obtain complete visual data, thereby accurately judging the changes of the dough during the whole kneading process, ensuring that the final dough meets the quality requirements. The final kneading image provides clear and intuitive image data for the final state of the dough, which is convenient for analyzing and judging whether it meets the expected fermentation standards. These final image data not only contribute to the final evaluation of the dough shape and quality, but also serve as the key input for subsequent learning and optimization. By combining the optimal dough state characteristics with image analysis, the system can quantitatively evaluate the final state of the dough. This quantitative evaluation is more accurate than traditional manual sensory evaluation and avoids human errors. The evaluation not only involves the dough shape, but also takes into account the intrinsic quality of the dough (such as elasticity, humidity, etc.), providing a comprehensive quality judgment. The final evaluation value provides a numerical value for further optimizing the kneading strategy. It is supported that if the evaluation value deviates from the target, the system can use this data to further adjust the parameters and improve the subsequent kneading process. The system adjusts its operating strategy based on the feedback after each kneading (the final dough state evaluation value) through a self-reinforcement learning mechanism. Each adjustment helps the system to continuously optimize and improve the kneading effect next time. The intelligent model can automatically adjust the kneading parameters, such as kneading time, kneading intensity and speed, eliminating the uncertainty in manual operation and further improving the accuracy and consistency of the kneading process. As the feedback data after each kneading is added to the learning model, the system can continuously accumulate experience, thereby optimizing the control strategy during long-term use, improving the kneading quality and production efficiency. The system can adjust its kneading strategy according to different dough types and environmental conditions, so that the consistency of dough quality can be maintained under various production environments.

[0023] In this specification, a system for selecting an intelligent dough kneading mode is provided, which is used to execute the method for selecting an intelligent dough kneading mode as described above, including: An image reconstruction module is used to obtain a real-time dough monitoring image in a dough kneading machine; perform edge sharpening and reconstruction on the real-time dough monitoring image, and perform time series frame sequence fitting to construct a dough monitoring time series frame sequence; The timing state diagram module is used to identify the dough contour frame by frame in the dough monitoring timing frame sequence, and to conduct in-depth mining of the dough timing state to construct a dough timing state diagram; The activity trend prediction module is used to mine the dynamic changes of yeast particles frame by frame and predict the activity trend at multiple moments in the dough monitoring time series frame sequence, so as to extract the optimal kneading time point; the dynamic kneading module is used to match the time state of the dough time series state diagram based on the optimal kneading time point, and then make dynamic kneading parameter decisions to generate dynamic kneading parameters; Parameter fine-tuning module, which is used to evaluate dough state based on dynamic kneading parameters, and to fine-tune and optimize instant kneading parameters, and to build an instant kneading fine-tuning strategy; Intelligent dough kneading control is used to perform comprehensive state evaluation based on the instant dough kneading fine-tuning strategy and self-reinforcement dough kneading performance learning to generate an intelligent dough kneading control model.

[0024] The present invention uses sensors such as cameras to obtain images of dough in a dough kneading machine in real time, provides raw data for subsequent analysis, performs edge sharpening processing on the real-time images obtained, highlights the contours and details of the dough, and ensures that subsequent image analysis is more accurate. Multiple frames of images are fitted in time-series frame order to construct a time-series frame sequence for dough monitoring, which helps the system identify dynamic changes of the dough during the kneading process, tracks contour changes of the dough in real time, and ensures that the system can detect slight changes in the dough morphology. After constructing a dough time-series state diagram, the change trend of the dough at different time points can be understood, providing accurate data for further analysis and decision-making. In-depth mining of the time-series state helps to identify potential kneading problems (such as over-kneading or over-relaxation), and improves the system's ability to identify and predict the dough state. By predicting yeast activity at multiple moments, the system can dynamically adjust parameters during the kneading process to ensure the optimal fermentation state of the yeast, accurately calculate the optimal kneading time point, and avoid kneading the dough when the yeast activity is not appropriate, thereby improving the fermentation effect and dough quality. The activity trend prediction module improves the control accuracy during the dough fermentation process. , ensure that the dough can reach the ideal fermentation state, ensure the accuracy of the kneading time by matching the optimal kneading time point, avoid over-kneading or under-kneading, dynamically adjust the kneading parameters according to the real-time feedback of the dough state, make the kneading process more personalized, and adapt to the needs of different doughs, accurate kneading control can improve the kneading efficiency and ensure that each stage of the dough is in the best state, by fine-tuning the kneading parameters, the system can respond to the changes in the dough state in time to ensure that the dough is always in the best state, no matter what changes occur in the dough during the kneading process, the system can flexibly adjust the kneading parameters to reduce the impact caused by environmental changes or other factors, the fine-tuning strategy can help maintain the consistency of the quality of each kneading and avoid quality fluctuations caused by operational errors, through self-reinforcement learning, the system can continuously accumulate experience and data, so as to continuously optimize the kneading control strategy, the generated intelligent kneading control model can automatically adapt to various production environments, improve the intelligence of the kneading process, and reduce manual intervention, with the accumulation of more data, the intelligent model will become more and more accurate, and continuously improve the kneading efficiency and product quality consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic diagram of the steps of a method for selecting an intelligent dough kneading mode according to the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2; Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0026] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0027] The present application example provides a method and system for selecting an intelligent dough kneading mode. The execution subject of the method and system for selecting an intelligent dough kneading mode includes but is not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of the present application, and the data processing platform includes but is not limited to: at least one of an audio image management system, an information management system, and a cloud data management system.

[0028] See also Figures 1 to 4 The present invention provides a method for selecting an intelligent dough kneading mode, and the method for selecting an intelligent dough kneading mode comprises the following steps: Step S1: obtaining a real-time dough monitoring image in a dough kneading machine; performing edge sharpening and reconstruction on the real-time dough monitoring image, and performing time series frame sequence fitting to construct a dough monitoring time series frame sequence; Step S2: Recognize the dough contour frame by frame on the dough monitoring time sequence frame sequence, and conduct deep mining of the dough time sequence state to construct a dough time sequence state diagram; Step S3: mining the dynamic changes of yeast particles frame by frame and predicting the activity trend at multiple moments in the dough monitoring time series frame sequence, so as to extract the optimal kneading time point; Step S4: performing time state matching on the dough timing state diagram based on the optimal dough kneading time point, and then making dynamic dough kneading parameter decisions to generate dynamic dough kneading parameters; Step S5: evaluating the dough state based on the dynamic kneading parameters, and fine-tuning and optimizing the instant kneading parameters to construct an instant kneading fine-tuning strategy; Step S6: Perform comprehensive state evaluation based on the instant dough kneading fine-tuning strategy, and conduct self-reinforcement dough kneading performance learning to generate an intelligent dough kneading control model.

[0029] The present invention provides a high-quality, real-time image data source for subsequent analysis through real-time visual monitoring of the dough state. Edge sharpening and reconstruction not only improves the contour recognition accuracy of the dough, but also reduces the problem of image quality affected by ambient light or clutter on the dough surface, thereby enhancing the accuracy of subsequent processing. Through sequential fitting of time series frames, dynamic tracking of dough changes during the kneading process can be achieved, providing a stable time series data basis for subsequent time series state analysis, identifying dough contours frame by frame, ensuring that every detail of the dough during the kneading process can be accurately captured, and providing high-precision data support for analyzing dough morphology changes. Sequence analysis can dig out subtle trends in dough changes, such as expansion, stretching, and rupture during fermentation, help monitor dough status in real time, and build a dough time-series state diagram, which not only improves the understanding of dough state changes, but also provides an important basis for subsequent time dynamic prediction and parameter optimization. Dynamic change mining makes it possible to monitor the activity of yeast in dough in real time, help accurately judge the key changes in the fermentation process, predict the changing trend of yeast activity, and provide a basis for formulating the best kneading time, avoiding excessive kneading or stopping too early, and ensuring that the dough enters the next step in the best state. By identifying the optimal kneading time point, Effectively avoid the phenomenon of over-kneading or under-kneading of dough due to improper time, improve dough quality, based on the precise matching of the optimal kneading time point and dough state, it can perform fine control at different fermentation stages of dough to improve the effect and quality of kneading. This step provides flexible decision support for kneading parameters, and can adjust the kneading intensity and time in real time to ensure that the dough is in the best state. Dynamic kneading parameter decision can better adapt to the actual state of the dough, avoid the potential impact of traditional fixed parameter control methods on dough quality, and improve the consistency and taste of the final product. By real-time monitoring of the dough state, the system can quickly respond to changes in the kneading process to ensure that the dough is always in the best state. Instant fine-tuning can make appropriate adjustments to the real-time state of the dough to avoid excessive or insufficient operation, thereby maximizing the quality of the dough. Through self-reinforcement learning, the system can optimize the control strategy according to the changing dough state, gradually accumulate experience, improve the intelligence level of kneading control, and realize personalized and customized kneading control, so that each batch of dough can get the most suitable kneading parameters, thereby greatly improving the consistency and quality of the product. As time goes by, the system will more and more accurately grasp the relationship between kneading parameters and dough quality, further improving production efficiency and reducing energy consumption.

[0030] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of the steps of a method for selecting an intelligent dough kneading mode of the present invention. In this example, the steps of the method include: Step S1: obtaining a real-time dough monitoring image in a dough kneading machine; performing edge sharpening and reconstruction on the real-time dough monitoring image, and performing time series frame sequence fitting to construct a dough monitoring time series frame sequence; In this embodiment, a suitable image capture device is selected, such as a high-definition camera or an industrial camera, to ensure that it can provide sufficiently clear images. The device needs to be able to operate stably in the working environment of the dough kneading machine. The camera is installed in an appropriate position to ensure that the image of the entire dough kneading area can be captured. During installation, the lighting conditions and the distance between the camera and the dough need to be considered. The VideoCapture class of the OpenCV library is used to start the camera and obtain real-time images. The cap.read() function can be used to periodically capture the current frame image to ensure the continuity and time consistency of image capture.

[0031] import cv2 cap = cv2.VideoCapture(0) # 0 means the default camera while True: ret, frame = cap.read() if not ret: break # Process the image (subsequent steps) Preprocess the captured original image, including conversion to grayscale, denoising, etc. Grayscale conversion can be implemented using the cv2.cvtColor() function, and denoising can be achieved using Gaussian blur. Apply the Canny edge detection algorithm to extract the edges in the image. Using the cv2.Canny() function, set a suitable threshold to accurately capture the edge contour of the dough. To enhance the edge effect, the edge image can be processed using the Laplacian operator or a custom sharpening filter. This makes the edge of the dough more obvious and facilitates subsequent analysis. gray =cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) blurred = cv2.GaussianBlur(gray, (5, 5), 0) edges = cv2.Canny(blurred, 100, 200) Store each frame of edge-sharpened images in a list so that a time-series frame sequence can be constructed later. You can define an empty list and append a new frame each time it is captured. Make sure that when storing time-series frames, they are stored in the order in which they were captured. You can mark the order of each frame by using a timestamp or a simple counter. Once a sufficient number of frames are captured, they can be organized into a time-series data structure to form a time-series frame sequence for subsequent analysis. frames = [] while True: ret, frame = cap.read() if not ret: break edges = process_frame(frame) # edge processing function frames.append(edges) # Store processed frames After completing real-time monitoring, make sure to release camera resources to avoid memory leaks. You can use cap.release() to release the camera. If necessary, you can save the time-series frame sequence to a file to facilitate subsequent data analysis and processing. You can use NumPy's np.save() function to save the array as a .npy file.

[0032] cap.release() np.save('dough_frames.npy', frames) # Save frame sequence.

[0033] Step S2: Recognize the dough contour frame by frame on the dough monitoring time sequence frame sequence, and conduct deep mining of the dough time sequence state to construct a dough time sequence state diagram; In this embodiment, necessary preprocessing steps are performed on each frame, including grayscale conversion, denoising and edge detection. Grayscale conversion uses cv2.cvtColor(), denoising generally uses Gaussian blur, and edge detection can use Canny algorithm.

[0034] import cv2 contours = [] for frame in frames: gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) blurred = cv2.GaussianBlur(gray, (5, 5), 0) edges = cv2.Canny(blurred, 100, 200) Use cv2.findContours() function to extract the contours in each frame. Extract the outer contour of the dough by setting the appropriate mode and method.

[0035] cnts, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) contours.append(cnts) # Store the contour of each frame Relevant features are calculated from the extracted contours, such as the area, perimeter, shape and center point of the contours. These features can reflect the state changes of the dough at different times.

[0036] features = [] for cnt in contours: area = cv2.contourArea(cnt) perimeter = cv2.arcLength(cnt, True) M = cv2.moments(cnt) centroid = (int(M['m10'] / M['m00']), int(M['m01'] / M['m00']))if M['m00'] != 0 else (0, 0) features.append({'area': area, 'perimeter': perimeter, 'centroid': centroid}) Arrange the extracted features in chronological order to form a time series feature dataset. You can use PandasDataFrame to store this data for easy analysis and visualization. Identify the changing trend of the dough state by analyzing the time series features. For example, you can calculate the rate of change, maximum value, minimum value, etc. of each feature to understand the fermentation and state change process of the dough. Choose a suitable visualization tool (such as Matplotlib) to draw a dough state diagram. Set the title, label, and color of the chart to clearly display the data.

[0037] import matplotlib.pyplot as plt plt.figure(figsize=(12, 6)) plt.plot(feature_df['area'], label='Area', color='blue') plt.plot(feature_df['perimeter'], label='Perimeter', color='orange') plt.title('Dough State Over Time') plt.xlabel('Time Frame') plt.ylabel('Feature Value') plt.legend() Step S3: mining the dynamic changes of yeast particles frame by frame and predicting the activity trend at multiple moments in the dough monitoring time series frame sequence, so as to extract the optimal kneading time point; In this embodiment, the dough monitoring time series frame sequence is analyzed frame by frame using image processing technology to extract yeast particles in each frame. A threshold segmentation method can be used to binarize the image to facilitate identification of yeast particles.

[0038] Use the cv2.threshold function to perform binarization and set an appropriate threshold to ensure that the yeast particles can be clearly separated. For each binarized image frame, use the cv2.findContours function to extract the outline of the yeast particles and calculate their number, area, and shape characteristics. Record the characteristics of the yeast particles in each frame, and use data structures (such as dictionaries or DataFrames) to store relevant information for each frame, including the number of particles, average area, and shape changes.

[0039] By comparing the yeast particle features of adjacent frames, dynamic change indicators such as the rate of change of the number of yeast particles and the rate of change of the area between each frame are calculated. These change indicators can help identify the dynamic trend of yeast activity. Using the extracted dynamic change data, an activity trend prediction model is constructed. Yeast activity can be modeled using linear regression, time series analysis, or machine learning methods such as random forests or support vector machines.

[0040] Take the extracted features (such as time, number of yeast particles, and area) as input, build a prediction model, and use historical data to train the model. For example, you can use the scikit-learn library to train the model: from sklearn.linear_model import LinearRegression model = LinearRegression() model.fit(X_train, y_train) # Training model.

[0041] Future time points are predicted to generate an activity trend graph reflecting the changes in yeast activity at different time points.

[0042] Based on the activity trend prediction results, the time point when the yeast activity is highest is identified. A threshold can be set to determine the peak position of the activity trend and extract the corresponding timestamp.

[0043] Find the local maximum in the activity trend graph and mark the optimal kneading time point. Peak detection can be achieved through the scipy.signal.find_peaks function: from scipy.signal import find_peaks peaks, _ = find_peaks(activity_trend, height=threshold) The obtained optimal kneading time point is associated with the actual kneading process record for subsequent operation and analysis.

[0044] The extracted optimal kneading time points and their corresponding yeast activity data are stored in the database for subsequent reference and analysis.

[0045] You can draw activity trend graphs, mark the optimal kneading time points, and use Matplotlib for visualization to help the production team understand the data intuitively.

[0046] Step S4: performing time state matching on the dough timing state diagram based on the optimal dough kneading time point, and then making dynamic dough kneading parameter decisions to generate dynamic dough kneading parameters; In this embodiment, data of the dough time series state diagram are extracted from the database. These data should include time series and state characteristics (such as temperature, humidity, physical properties of the dough, etc.).

[0047] Ensure the determination of the optimal kneading time point and record its corresponding state characteristics for matching and analysis.

[0048] Time state matching: Determine the optimal kneading time point and format it into a matching data format (such as a timestamp).

[0049] Use the conditional screening method to extract the state feature data corresponding to the optimal kneading time point from the dough time series state diagram. You can use the loc method in the Pandas library for accurate matching: matched_state = state_data.loc[state_data['time'] == optimal_time], and record the matching results, including the state features of the optimal time point and its corresponding time information. Based on the extracted state feature data, establish a dynamic kneading parameter decision model. You can use rule-based decision logic or machine learning models (such as decision trees and random forests) to make parameter decisions.

[0050] Set up decision rules, such as: If the temperature is above a certain threshold, the kneading time is increased.

[0051] If the dough humidity is too low, increase the amount of water. Input the matched state feature data into the decision model to calculate the dynamic kneading parameters that need to be adjusted. These parameters should include kneading time, kneading intensity, speed, etc. Adjust the parameters according to the recommendations output by the model: if matched_state['temperature'] > threshold: dynamic_parameters['kneading_time'] += additional_time The generated dynamic kneading parameters are saved in the database, and the corresponding optimal kneading time points and state characteristics are recorded for subsequent analysis and monitoring. During the actual kneading process, the state changes of the dough are monitored in real time, and the dynamic parameters are adjusted according to the feedback to ensure the optimal state of the dough.

[0052] Step S5: evaluating the dough state based on the dynamic kneading parameters, and fine-tuning and optimizing the instant kneading parameters to construct an instant kneading fine-tuning strategy; In this embodiment, in actual operation, the dynamic kneading parameters generated in the previous step are applied to the dough kneading machine to ensure that the equipment can be adjusted according to these parameters, covering key parameters such as kneading time, speed, and intensity. During the implementation process, the operating status of the equipment is monitored in real time to ensure that it works normally according to the set parameters. The industrial control system or PLC (programmable logic controller) can be used to set the parameters. After the dynamic parameters are applied, the state changes of the dough are continuously monitored, focusing on indicators such as the temperature, humidity, texture and appearance of the dough. These data can be captured by sensors (such as temperature and humidity sensors, texture measuring instruments), and the dough state is recorded regularly using a data acquisition system for subsequent analysis. The data can be stored in a database to ensure the integrity and traceability of the data. A dough state evaluation model is constructed, and the dough state data collected previously is used to define evaluation criteria. The evaluation criteria may include the uniformity, viscosity, elasticity, etc. of the dough. A machine learning algorithm (such as a support vector machine or a decision tree) can be used to train the model to perform state evaluation based on real-time data. According to the model evaluation results, the dough is judged. It can be judged whether the current dough state meets the expected goal. If the dough state is not ideal (such as too dry or wet), it is necessary to make immediate parameter fine-tuning. For example, if the dough is too sticky, the kneading time can be reduced or the kneading intensity can be reduced. If the dough is too dry, the amount of water added can be increased. The feedback control mechanism is used to dynamically adjust the kneading parameters according to the evaluation results, and the fine-tuned parameters are recorded in the database to ensure that all adjustments are traceable. The records should include the reasons for fine-tuning, the adjusted parameter values ​​and their corresponding timestamps. The immediate fine-tuning strategy is evaluated and the effect after adjustment is analyzed. If the effect is significant, it can be included in the standard operating procedure. If the effect is not good, the strategy needs to be further adjusted, historical data should be reviewed regularly, and the fine-tuning strategy should be optimized to improve decision-making efficiency. During the whole process, the changes in dough state and kneading parameters are continuously monitored, feedback data is collected, and these data are used for model retraining and optimization. Through machine learning methods, the evaluation model and fine-tuning strategy are continuously updated to adapt to the characteristics and changes of different batches of dough to ensure that each kneading process can achieve the best effect.

[0053] Step S6: Perform comprehensive state evaluation based on the instant dough kneading fine-tuning strategy, and conduct self-reinforcement dough kneading performance learning to generate an intelligent dough kneading control model.

[0054] In this embodiment, a comprehensive state evaluation model is constructed based on the real-time kneading fine-tuning. The model should take into account multiple factors, including the physical properties of the dough (such as temperature, humidity, elasticity), yeast activity and historical kneading parameters.

[0055] Select appropriate evaluation indicators, such as dough viscosity, uniformity, and fermentation degree, and use methods such as multivariate linear regression or decision trees for modeling. You can use the scikit-learn library to create and train the model: fromsklearn.linear_model import LinearRegression model = LinearRegression() model.fit(X_train, y_train) # training model In actual operation, the dough state data is continuously collected, including real-time monitoring of temperature, humidity, kneading parameters and their corresponding evaluation results. Use sensors and data recording systems to automate data collection. Extract feature data, which should include the impact of instant fine-tuning parameters on dough state, historical state evaluation results, etc. Store these data in a database for subsequent analysis and model training. Design a self-reinforcement learning framework and choose a suitable algorithm, such as Q-learning or deep Q network (DQN). This framework should be able to automatically adjust the kneading parameters based on the dough state evaluation results to optimize the kneading process. Determine the state space (such as the current state characteristics of the dough), the action space (adjustable kneading parameters) and the reward mechanism (such as dough quality score). The reward mechanism should be designed reasonably to encourage the learning algorithm to optimize the kneading process. For example, a scoring system based on the final dough quality can be set up: Use the collected data and the self-reinforcement learning algorithm to train the model. Through continuous trial and error and feedback, the algorithm will learn the relationship between the optimal kneading parameters and dough state. During the training process, real-time feedback is used to adjust the parameters of the model to ensure that the model can adapt to changes in the dough state. Use the experience replay mechanism to store historical states to improve learning efficiency and effectiveness.

[0056] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Step S11: obtaining a real-time dough monitoring image in a dough kneading machine; Step S12: performing edge sharpening and reconstruction on the real-time dough monitoring image to construct a detail-optimized and reconstructed dough image; Step S13: extracting frame-by-frame image changes of the detail-optimized and reconstructed dough image to obtain a dough monitoring image of each frame; Step S14: Calculate the time stamp of sequential frame acquisition for the detail optimized and reconstructed dough image to obtain the acquisition time stamp of each frame; Step S15: performing time-series frame sequence fitting on each frame of the dough monitoring image according to the acquisition timestamp of each frame, and constructing a dough monitoring time-series frame sequence.

[0057] In this embodiment, a suitable camera (such as an industrial camera or a high-resolution USB camera) is selected to ensure that a clear image can be provided. The camera is installed at an appropriate position of the dough kneading machine to ensure that the entire dough kneading area can be covered. Use image capture software (such as OpenCV or a custom image acquisition program) to set camera parameters (such as resolution, frame rate, exposure, etc.). Ensure that the camera connection is stable and can transmit image data in real time. Start the camera and start capturing dough images in real time. Save the captured image data stream to the memory or directly transfer it to the processing module for subsequent processing. Perform color space conversion (such as from BGR to grayscale) on the acquired real-time dough monitoring image to simplify processing. Select a suitable edge detection algorithm, such as Canny edge detection or Sobel operator. Apply the selected edge detection algorithm to extract edge information in the image. Use a sharpening filter (such as a Laplacian filter) to enhance the edge features of the image. Synthesize the processed edge information with the original image to generate a dough image with optimized detail reconstruction. Save the processed image for subsequent frame-by-frame extraction. Load the sequence of dough images with optimized detail reconstruction into the memory. Compare each image frame with the previous one and extract the changes using a difference detection algorithm (e.g., absolute difference method). Thresholds can be set to filter out small or unimportant changes. Save the extracted changes as new image frames to form a new dough monitoring image sequence. Ensure that each frame of the change image reflects the dynamic changes in the dough state. Record the initial time at the beginning of image capture (e.g., using Python's time module). For each image frame, calculate the difference from the initial time to generate a timestamp. Record the acquisition timestamp in milliseconds or seconds in a list. Load the acquisition timestamp and corresponding image frame of each frame from storage. Sort the image frames according to the timestamps to ensure that the image sequence is arranged in chronological order. This can be achieved using the sorted() function in Python or other sorting algorithms. The sorted image frames and timestamps form a time-series frame sequence for subsequent analysis. Save the time-series frame sequence as a video file or image sequence for visualization and analysis.

[0058] In this embodiment, the specific steps of step S12 are: Perform pixel-by-pixel recognition on the real-time dough monitoring image to extract all pixel points in the image; Perform global pixel average calculation on all pixels in the image to generate the global pixel average of the image; Image noise suppression is performed according to the global pixel average value of the image, thereby obtaining a denoised and optimized dough monitoring image; Perform histogram equalization on the denoised and optimized dough monitoring image to generate an image grayscale histogram; Perform grayscale distribution stretching on the image grayscale histogram and calculate the grayscale value of the stretched pixel; performing contrast enhancement according to the grayscale values ​​of the stretched pixels to obtain a contrast enhanced dough monitoring image; Perform edge gradient information analysis on the texture detail enhanced dough monitoring image to obtain an edge map; Performing slight texture change sharpening on the edge map to generate a slight texture change sharpened edge map; The contrast enhanced dough monitoring image is reconstructed by edge sharpening according to the micro-texture change sharp edge map to construct a detail optimized reconstructed dough image.

[0059] In this example, an image processing library (such as OpenCV, PIL, or scikit-image) is used to load a real-time dough monitoring image. Make sure the image format is correct, usually an RGB or grayscale image. Convert the image to an array for pixel-by-pixel access. In Python, the numpy library can be used to convert the image data into an array with a shape of (height, width, number of channels). Iterate over each pixel and extract the RGB value (or grayscale value) of each pixel through a double loop (outer loop traverses rows and inner loop traverses columns). Create a list or array to store the extracted pixel values ​​for subsequent processing. The value of each pixel can be stored in a tuple or dictionary, containing its coordinates (x, y) and the corresponding color value. Store the data of all pixels in memory or write them to a temporary file for subsequent processing. Record the total number of extracted pixels and basic information of the image (such as width, height) to provide context for subsequent analysis. Create variables to store the sum of each color channel (e.g., total_red, total_green, total_blue) and pixel count variables, initialized to zero. Traverse the extracted pixel list and add the RGB value of each pixel to the corresponding sum variable. Each time a pixel is traversed, the pixel count variable is increased by one. After the traversal is completed, the accumulated sum is divided by the total number of pixels to calculate the average value of each channel: average_red = total_red / pixel_count, average_green = total_green / pixel_count, average_blue = total_blue / pixel_count, stores the calculated global average as a dictionary or tuple in the format of (average_red, average_green, average_blue). Record the type of average (such as RGB or grayscale) to provide necessary information for subsequent image processing. Determine the noise type, common ones are Gaussian noise, salt and pepper noise, etc. Select an appropriate denoising algorithm based on the noise type, such as mean filtering, median filtering, or Gaussian filtering. Apply the selected filter to the image. Take Gaussian filtering as an example: use the cv2.GaussianBlur function and set an appropriate convolution kernel size (such as 5x5 or 7x7) to smooth the image. You can also implement a custom denoising algorithm, using convolution operations to traverse the image and calculate the weighted average of each pixel. Generate a denoised image, retaining the same dimensions and data types as the original image to ensure compatibility with subsequent processing. Store the denoised and optimized image in memory or write it to a file to prepare for subsequent histogram equalization and other processing steps. If the current image is an RGB image, it needs to be converted to a grayscale image for histogram equalization. Convert the image from RGB to grayscale using the cv2.cvtColor function. Use the cv2.calcHist function to calculate the histogram of the grayscale image and return the number of pixels at each grayscale level. The range of the histogram is usually 0 to 255, corresponding to all possible values ​​of an 8-bit grayscale image. Apply the cv2.equalizeHist function to equalize the grayscale image, which adjusts the grayscale distribution of the image to make the contrast of the image more uniform. Store the equalized image in memory and save the histogram data for subsequent analysis and visualization. You can choose to plot and display the histogram to facilitate observation of the equalization effect. Find the maximum and minimum grayscale values ​​in the equalized image, which can be achieved by traversing all pixels of the image. Stretch the grayscale value of each pixel using the following formula: new_value = (pixel_value−min_value) / (max_value−min_value) ×255, linearly stretching the grayscale value to the range of 0 to 255. Apply the calculated new grayscale value to the image to generate an updated image.

[0060] Select the contrast enhancement method according to your needs. Common methods include linear contrast enhancement, logarithmic transformation, or gamma correction. For linear enhancement, you can do the following: Apply the linear transformation formula to each pixel: enhanced_value=α×pixel_value+β, where α is the contrast gain coefficient and β is the brightness offset. Make sure that the enhanced pixel value is between 0 and 255. Apply a clipping operation if necessary to set the out-of-range value to 0 or 255. Select a suitable edge detection algorithm, such as the Sobel operator, Canny edge detection, or Laplace operator. Canny edge detection usually works well and can effectively handle noise and extract details. If you choose the Canny algorithm, use the cv2.Canny function, pass the image in, and set a suitable threshold (such as a low threshold and a high threshold). If you use the Sobel operator, first smooth the image, then calculate the gradients in the x and y directions, and finally synthesize the edge map. Generate an edge map to display the main edge information extracted from the image. It is usually a binary image, with white representing edges and black representing non-edges. Store the edge map in memory for subsequent sharpening of small texture changes. Select a suitable sharpening filter, such as a Laplacian filter or a custom sharpening convolution kernel. Use the cv2.filter2D function to apply the selected sharpening convolution kernel to the edge map to enhance edges and details.

[0061] Ensure that the sharpening operation does not introduce too much noise, and combine it with a smoothing operation if necessary. Generate a micro-texture change sharpened edge map to show enhanced details and clarity. Store the sharpened micro-texture change edge map in memory for subsequent detail optimization and reconstruction processing. Combine the micro-texture change sharpened edge map with the contrast enhanced image using the weighted superposition method: optimized_image=enhanced_image+λ×sharpened_edges. Among them, λ is the weight coefficient that controls the degree of edge enhancement. Normalize pixel values: Ensure that the pixel values ​​of the reconstructed image are between 0 and 255, and perform necessary cropping operations. Generate a detail optimized and reconstructed dough image to facilitate the observation of the details and state of the dough. Store the final detail optimized and reconstructed dough image in a database or file system and prepare it for display and analysis.

[0062] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Step S21: performing dough contour recognition frame by frame on the dough monitoring time series frame sequence, and extracting the dough contour line of each frame; Step S22: Tracking the dynamic changes of the dough contour line of each frame to obtain a dynamic change trajectory of the dough contour; Step S23: performing contour change trend analysis on the dynamic change trajectory of the dough contour to generate dough contour change trend characteristics; Step S24: evaluating the dough surface roughness of the dough monitoring time series frame sequence to generate a dough surface roughness time series curve; Step S25: performing frame-by-frame analysis of the dough crack generation situation on the dough monitoring time series frame sequence to obtain crack evolution characteristics in the dough; Step S26: Conduct in-depth exploration of the dough time-series state based on the dough surface roughness time-series curve, dough contour change trend characteristics, and dough crack evolution characteristics to construct a dough time-series state diagram.

[0063] In this embodiment, an image processing library (such as OpenCV) is used to load the image sequence of each frame to ensure that the image format is consistent, such as RGB or grayscale images, and the image is preprocessed, including denoising, smoothing and grayscale. Common denoising methods include Gaussian filtering to reduce the influence of noise in the image, smoothing the image using the cv2.GaussianBlur function, and converting it to a grayscale image using cv2.cvtColor. An edge detection algorithm, such as Canny edge detection, is applied to identify the edges in the image, and a suitable threshold is set using the cv2.Canny function to effectively extract The contour of the dough is extracted from the edge map using the cv2.findContours function, which returns a list of coordinate points of the contour. The outermost contour can be selected to represent the contour of the dough. The extracted contour is stored in a list for subsequent processing. The contour information of each frame is saved to a file or database, and the contour point data of each frame and its corresponding timestamp are recorded to provide basic data for dynamic change tracking. For continuous contours in time series frames, shape matching algorithms (such as Hu moments or shape contexts) are used for matching. The dynamic changes of contours are judged by calculating the similarity between contours. cv2.The matchShapes function obtains the similarity score between contours, compares contours frame by frame, records the changes between each frame and the previous frame, calculates the position, area, and shape changes of the contours, stores the change information (such as the moving distance and shape changes of the contours) in an array, forms a dynamic change trajectory, and uses visualization tools (such as Matplotlib) to draw a dynamic change trajectory diagram to show the changes of the dough contour over time. Different colors or marks can be used to represent different degrees of change to help analyze dynamic trends. The dynamic change trajectory is stored in a database, and the change parameters at each time point are recorded for subsequent analysis and feature extraction. Trend modeling is performed on the dynamic change trajectory data. Linear regression or moving average method can be used to analyze the change trend of the dough contour. Linear regression modeling is performed using Python's scikit-learn library to obtain a functional expression of the change trend. Important features such as the rate of change and maximum change amplitude are extracted from the model. These features can help understand the changes in the dough state, calculate the standard deviation and mean of the changes, analyze its stability and fluctuations, and use visualization tools. Draw a trend graph to show the changing characteristics of the dough profile for easy analysis and interpretation. You can choose different time periods for comparison and analyze the differences in characteristics. Store the extracted trend characteristics in the database and record the characteristic values ​​and their corresponding time for subsequent analysis and report generation. Select a suitable roughness evaluation method, such as average roughness (Ra) or root mean square roughness (Rq). These indicators can be extracted from the grayscale value changes of the image. Calculate the grayscale value of each frame of the image and calculate the roughness. You can use the numpy library to operate arrays and quickly calculate the roughness index. Record the roughness value of each frame and generate a timing curve. Use Matplotlib to draw a timing curve graph to show the trend of roughness over time. Save the roughness timing curve and its values ​​in the database. Record the roughness value at each time point for subsequent analysis and reporting. Select a suitable crack detection algorithm, such as a method based on edge or texture features. You can use Canny edge detection or Hough transform to detect cracks. Apply the crack detection algorithm to each frame of the image to identify and extract crack information. Use cv2.The findContours function extracts the contour of the crack, compares the changes of the crack frame by frame, and records the number, length, width and other characteristics of the crack. The evolution of the crack can be analyzed by calculating the total length and number of cracks in each frame. The crack evolution characteristics are stored in the database, and the crack characteristics at each time point are recorded for subsequent analysis and report generation. The dough surface roughness time series data, contour change trend characteristics and crack evolution characteristics are extracted from the database and integrated into a comprehensive data set. The appropriate state analysis model can be selected. Multiple regression analysis or cluster analysis methods can be used to deeply explore the time series state of the dough. The integrated data can be deeply mined to analyze the relationship between different characteristics and find out the key factors affecting the dough quality. The dough time series state diagram is constructed based on the analysis results to show the characteristic changes of the dough under different states. The time series state diagram is drawn using visualization tools to show the state change trend of the dough for intuitive understanding and analysis. The time series state diagram is stored in the database, and relevant reports are generated to record the analysis results and suggestions to support decision-making and subsequent research. .

[0064] In this embodiment, refer to Figure 4 , is a flowchart of detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Step S31: mining the dynamic changes of yeast particles frame by frame in the dough monitoring time series frame sequence to extract the dynamic change characteristics of yeast particles; Step S32: performing yeast temporal activity analysis according to the dynamic change characteristics of yeast particles to generate a yeast temporal activity curve; Step S33: Evolving the activity variation law of the yeast time series activity curve to obtain the yeast activity variation law; Step S34: performing multi-time activity trend prediction based on the yeast activity change rule, thereby generating yeast activity trend prediction data at multiple times; Step S35: Calculate the optimal kneading time based on the yeast activity trend prediction data at multiple moments, so as to extract the optimal kneading time point.

[0065] In this embodiment, an image processing library (such as OpenCV) is used to load the time-series image frames of dough monitoring to ensure that the loading format of each frame is consistent and can be easily processed. Each frame of the image is preprocessed, including denoising and contrast enhancement. Gaussian filtering (cv2.GaussianBlur) can be used to reduce image noise, and histogram equalization can be used to improve contrast. The image is converted into a grayscale image for subsequent feature extraction. Morphological operations (such as opening and closing operations) are applied to enhance the visualization of yeast particles and segmentation is performed. The cv2.morphologyEx function is used for morphological processing, and threshold segmentation (such as the Otsu method) is used to extract yeast particles. The mother particles are extracted and the particle contours in each frame are extracted using the cv2.findContours function. The number, average area, shape features (such as aspect ratio, roundness, etc.) of yeast particles in each frame and the movement trajectory of the particles are calculated. These features can be calculated by traversing the extracted contours. The extracted features are stored in a data structure for subsequent analysis. Indicators for evaluating yeast activity, such as the number of particles, average area, or particle movement speed, are determined. These indicators can well reflect the changes in yeast activity. For each frame of the image, the activity value is calculated based on the selected activity indicator. For example, the number of particles is used as the activity value, or the average area of ​​the particles is normalized to the activity score. The calculated activity The values ​​are stored in a time series data structure. Visualization tools such as Matplotlib are used to draw yeast time series activity curves, with time on the x-axis and activity values ​​on the y-axis. To ensure that the curve is clear and easy to read, key time points (such as when yeast activity changes significantly) can be marked in the graph for easy analysis. The generated yeast time series activity curve and its values ​​are saved to the database, and the activity value at each time point is recorded for subsequent analysis and reporting. In order to analyze the changing pattern, the yeast time series activity curve is first smoothed. The moving average method or Savitzky-Golay filter can be used to reduce the impact of noise on the regularity analysis. The rolling function of the pandas library is used to implement the moving average. Set an appropriate window size, observe the smoothed activity curve, and use linear regression or polynomial regression to analyze the trend of activity changes. You can use the regression model in scikit-learn to build a model and extract key features, such as slope and inflection point position, to characterize the trend of activity changes. Based on the analysis results, summarize the law of yeast activity changes, including the time period and amplitude of activity enhancement and reduction. These laws help understand the behavior of yeast during fermentation and select a suitable prediction model, such as ARIMA model, time series regression or LSTM (long short-term memory network). Select the optimal model based on the data characteristics. If the data volume is small, you can consider using a linear regression model.If the data volume is large and has complex nonlinear relationships, you can consider the LSTM model, divide the historical activity data into a training set and a test set, use the training set to train the prediction model, use the scikit-learn or statsmodels library to train the model, adjust the model parameters, ensure that the model can accurately capture the activity change law, use the trained model to predict the future yeast activity, set the prediction time point, generate activity trend prediction data at multiple moments, visualize the prediction results, draw the future activity trend curve, and compare it with the historical data to observe the rationality of the prediction and determine the standard for the optimal kneading time, such as when the yeast activity value reaches the maximum. When the yeast activity value is large, or the activity value remains at a stable high level, the yeast activity trend prediction data is analyzed to identify the peak position and duration of the activity value. The argmax function in the numpy library can be used to find the index of the maximum activity value, and the time point when the activity value reaches a preset threshold is calculated as the potential kneading time. The optimal time point is selected according to the activity change law and historical data to ensure that the dough is kneaded when the yeast activity is the highest and the fermentation effect is the best. Multiple time points can be considered for flexible selection in actual operation. The extracted optimal kneading time point is stored in the database, and its corresponding activity value and timestamp are recorded to provide a reference for actual production. ;

[0066] In this embodiment, the specific steps of step S35 are: Perform multi-time activity trend peak calculation on yeast activity trend prediction data at multiple times, and extract the activity trend peak of each time period; The activity trend peak in each time period is used to identify the inflection point of activity change and marked as the highest peak point of yeast activity; Define preset kneading time parameters; Calculate the kneading buffer time at the highest peak point of yeast activity according to the preset kneading time parameters to obtain the kneading buffer time at the highest peak point; The optimal kneading time is calculated based on the kneading buffer time at the highest peak point to obtain the optimal kneading time point.

[0067] In this example, yeast activity trend prediction data is extracted from the database. The data should include time series and corresponding activity values. The data is organized into a format suitable for analysis, such as converting it into a DataFrame using the Pandas library. A suitable peak detection algorithm is selected, such as the find_peaks function in the SciPy library, which can effectively identify local peaks in the signal. The peak detection parameters are set, including the minimum peak height and the minimum distance, to ensure that only significant activity peaks are detected. The scipy.signal.The find_peaks function analyzes the activity trend data, identifies all peaks and their corresponding time points and activity values, stores the detected peaks in a list or DataFrame, records the time, activity value and position of each peak, sorts the extracted peaks, and presents them visually to facilitate observation of the activity trend peaks in each time period. In the visualization graph, you can use Matplotlib to draw the activity curve and mark the peaks, save the peak results to a file or database, and provide basic data for subsequent activity change inflection point identification. Select a suitable inflection point detection algorithm, such as using the first-order derivative or rate of change. You can identify the rising and falling inflection points by calculating the derivative of the activity value, set a threshold, and determine the obvious inflection point location. For example, by setting a threshold for the minimum rate of change, calculate the first-order derivative of the activity value sequence, and identify the point where the positive changes to the negative as the inflection point. In Python, you can use the diff function of the numpy library to calculate the derivative. For each peak, check the changes before and after it to identify Identify the highest peak of the activity value, mark the identified highest peak point as "the highest peak point of yeast activity", and record its timestamp and activity value. Store all inflection points and their corresponding peak information in the database for subsequent use. Select appropriate kneading time parameters based on actual operation experience or experimental data. This time should take into account the fermentation characteristics of yeast and the physical properties of dough. For example, the kneading time can be set to 30 minutes, 45 minutes or 60 minutes, etc. The specific time period can be adjusted according to process requirements. Store the selected kneading time parameters in the configuration file or database for calling in subsequent steps. It can be defined as a global variable for reference in subsequent calculations. Confirm whether the preset kneading time parameters meet the production standards and record the basis for their selection for subsequent evaluation and adjustment. Define the calculation formula for the buffer time. For example, the buffer time can be set to "the time of the highest peak point of yeast" plus "the preset kneading time". For example, if the highest peak point occurs at time t and the preset kneading time is 30 minutes, then the buffer time is t + 30 minutes, extract the time from the highest peak point of the mark, combine it with the preset kneading time, calculate it, ensure that the calculation result of the buffer time is logical, and combine the buffer time result with the current time within a reasonable time range to determine the optimal kneading time point. For example, if the buffer time is t, the optimal kneading time is the current time plus t. You can use Python's date and time processing library (such as datetime) for time calculation to verify whether the calculated optimal kneading time point is within a reasonable time range and meets the requirements of production scheduling, ensuring that the time point can maximize the yeast activity and improve the fermentation effect of the dough. The optimal kneading time point is recorded in the database, and relevant information (such as the corresponding yeast activity value, timestamp, etc.) is recorded for subsequent query, and a report is generated to output the optimal kneading time point and its suggestions for reference by the production team. .

[0068] In this embodiment, step S4 includes the following steps: Step S41: performing time state matching on the dough time sequence state diagram based on the optimal kneading time point, and extracting the dough time sequence state characteristics at the optimal kneading time point; Step S42: Analyze the optimal dough state of the dough time sequence state diagram to generate optimal dough state characteristics; Step S43: performing an optimal state deviation analysis on the optimal dough state characteristics based on the dough temporal state characteristics at the optimal dough kneading time point to obtain the dough state deviation characteristics at the optimal dough kneading time point; Step S44: making dynamic kneading parameter decisions based on the dough state deviation characteristics at the optimal kneading time point to generate dynamic kneading parameters.

[0069] In this embodiment, the data of the dough time series state diagram is extracted from the database, including time series, state features (such as roughness, contour changes, crack conditions, etc.). Ensure that the data format is suitable for processing, such as using Pandas DataFrame to facilitate subsequent operations. Determine the optimal kneading time point and convert it into a suitable format, such as a timestamp. Use conditional filtering to extract the state features corresponding to the optimal kneading time point from the time series state data. The loc method of Pandas can be used to achieve accurate matching. Extract relevant dough state features from the matching results, including but not limited to surface roughness, dynamic change trajectory and crack conditions. These features can be obtained through simple indexing. Store the extracted state features in a new data structure for subsequent analysis. Save the extracted dough time series state features of the optimal kneading time point to a database or file, record its corresponding timestamp and feature value, and provide basic data for subsequent analysis. Determine the analysis method, and use descriptive statistical analysis, cluster analysis, or principal component analysis (PCA) to extract the overall state features of the dough. Select a suitable Python library, such as scikit-learn for cluster analysis and statsmodels for statistical analysis. Preprocess the data of the dough time series state diagram to ensure the integrity and consistency of the data. Deal with missing values, outliers and other issues to ensure the reliability of the analysis results. You can use the dropna method of Pandas to remove missing values ​​and perform data cleaning. Extract the optimal state characteristics of the dough according to the selected analysis method. For example, if PCA is used, you can project the data into a low-dimensional space and extract the main components. Record the contribution rate of each feature to understand which features play an important role in the overall state. Determine the method for calculating the deviation. You can use indicators such as absolute error, relative error or standard deviation to quantify the deviation of the state characteristics. Select an appropriate Python library for calculation, such as numpy for numerical calculation and scipy for statistical analysis. Extract the dough time series state characteristics and the optimal dough state characteristics at the optimal kneading time point from the database to ensure that the data format is consistent. Ensure that the features of the two sets of data match and set the same feature dimension for comparison. For each feature, calculate the deviation between the optimal kneading time point feature and the optimal dough state feature. For example, use the following formula to calculate the absolute deviation: Deviation = actual state feature − optimal state feature The decision method for determining the dynamic kneading parameters can use methods such as rule engines, linear regression or fuzzy logic. Choose a suitable Python library, such as sklearn for linear regression and fuzzywuzzy for fuzzy logic decision making. Establish a decision model based on the dough state deviation characteristics. For example, establish a linear regression model to predict the relationship between the optimal kneading time and state deviation. Use the deviation characteristics as input variables and dynamic parameters (such as kneading time and kneading intensity) as output variables for model training.Use the established model to predict the current dough state deviation and generate the corresponding dynamic kneading parameters. For example, input the current state deviation and output the recommended kneading time and intensity. Ensure that the generated dynamic parameters are reasonable and executable and meet the actual operation requirements.

[0070] In this embodiment, step S5 includes the following steps: Step S51: performing real-time dough kneading control based on dynamic dough kneading parameters, and acquiring real-time dough kneading monitoring images; Step S52: performing dough shape analysis on the real-time dough kneading monitoring image to obtain the real-time dough kneading shape; Step S53: performing dough state evaluation on the real-time dough shape during kneading to generate a real-time dough state evaluation value; Step S54: Based on the real-time kneading state evaluation value, the dynamic kneading parameters are fine-tuned and optimized to construct an instant kneading fine-tuning strategy.

[0071] In this embodiment, real-time dynamic dough kneading parameters are extracted from the database. These parameters should include dough kneading time, dough kneading speed, dough kneading intensity, etc. These parameters should be adjusted according to the analysis results of the previous stage, and the initial state of the control system is set to ensure that all control devices (such as dough kneading machines) can receive and execute these dynamic parameters. A real-time control system is constructed, and a PLC (programmable logic controller) or an industrial computer is used to control the operation of the dough kneading machine. Real-time control of the dough kneading machine is achieved through programming to ensure that it can be adjusted according to the dynamic parameters. A suitable control algorithm (such as PID control) is used to ensure accurate control of dough kneading, and the dough kneading intensity and time are adjusted in time. A high-resolution camera is installed on the dough kneading machine to obtain real-time monitoring images of the dough, and the position of the camera is ensured to clearly capture the shape and state of the dough. Image acquisition software is configured to regularly capture real-time images and store them locally or upload them to a cloud server for subsequent analysis. In the real-time control system, the dynamic parameters and corresponding monitoring images at each time point are recorded, and the data are saved in a database or file system for subsequent analysis and verification. The monitoring system should have a real-time feedback function to ensure that the operation can be adjusted in time when the parameters change. The real-time monitoring image is preprocessed using an image processing library (such as OpenCV), including noise removal and contrast enhancement. Gaussian filtering and histogram equalization can be used to improve image quality. The image is converted into a grayscale image for subsequent processing and analysis. The Canny edge detection algorithm is applied to identify the edges in the image. The edge map is obtained through the cv2.Canny function. The cv2.findContours function is used to extract the contour of the dough, obtain the boundary information of the dough, and calculate the shape characteristics of the dough, including the area, perimeter, aspect ratio, etc. of the dough. These features will be used for subsequent state evaluation, using cv2.contourArea and cv2.The arcLength function obtains the area and perimeter of the contour, and then calculates shape features such as the aspect ratio. The extracted dough shape features are saved in the database, and the corresponding timestamp and image information are recorded for subsequent analysis and decision support. The criteria for dough state evaluation are determined. For example, shape features (such as area, aspect ratio, uniformity, etc.) are used to determine whether the dough state is qualified. Thresholds can be set to define the standards for good and bad states for subsequent evaluation. According to the defined evaluation criteria, a dough state evaluation model is constructed. A simple rule engine can be used to generate state evaluation values ​​through logical judgment. For example, if the aspect ratio of the dough is within a certain range, the dough state is considered to be good, otherwise it is considered to be bad. The dough shape features of each frame are evaluated and the state evaluation value is calculated. The extracted shape features are input into the model to obtain real-time evaluation results, and the evaluation values ​​are recorded and associated with the corresponding timestamps and image information. According to the real-time state evaluation values, set A fine-tuning strategy can be designed. For example, if the evaluation value is lower than the preset threshold, the kneading time or the kneading intensity can be increased; if the evaluation value is higher than the threshold, the kneading time may need to be reduced. A clear set of rules can be formulated to guide the adjustment of dynamic parameters. The state evaluation value can be monitored in real time, and the kneading parameters can be dynamically adjusted according to the fine-tuning strategy. The control parameters can be directly modified through the control system interface or API interface. The feedback control algorithm can be used to iteratively adjust the parameters according to the current state and preset goals. After adjusting the parameters, the state changes of the dough are continuously monitored to evaluate the effectiveness of the new parameters. The effect of the fine-tuning strategy can be verified by comparing the state evaluation values ​​before and after the adjustment. A short-term feedback cycle can be set, such as re-evaluating the state every few minutes to achieve timely adjustment. The adjusted dynamic kneading parameters and their corresponding state evaluations are recorded in the database for subsequent analysis and optimization. Historical data can be reviewed regularly to evaluate the effectiveness of the fine-tuning strategy, and the adjustment plan can be continuously optimized according to actual conditions. .

[0072] In this embodiment, step S6 includes the following steps: Step S61: extracting a final dough kneading image according to the real-time dough kneading monitoring image; Step S62: performing a comprehensive state evaluation on the final dough kneading image according to the optimal dough state characteristics to obtain a final dough state evaluation value; Step S63: Based on the final dough state evaluation value, the instant dough kneading fine-tuning strategy is self-reinforced to learn the dough kneading performance, thereby generating an intelligent dough kneading control model.

[0073] In this embodiment, the real-time image capture system is ensured to work properly. The camera should be properly installed and be able to capture high-quality images at different dough kneading stages. The image acquisition frequency is configured to ensure that clear images can be obtained at critical moments. Usually, the image can be captured every few seconds. The captured real-time image is processed using an image processing library (such as OpenCV), such as denoising, contrast enhancement, and conversion to grayscale. During the processing, the clarity and details of the image are ensured to be retained for subsequent status evaluation. A trigger condition is set, such as when the dough kneading time reaches a predetermined value or the dough state meets the standard, the current image is extracted as the final dough kneading image. The trigger mechanism is implemented in the code using a conditional judgment statement (such as an if statement). Ensure that the final image is automatically saved when the conditions are met. Save the extracted final kneading image to local storage or the cloud for subsequent access and analysis. You can use the save method of the PIL library to save the image in JPEG or PNG format. Record the timestamp and related parameters of the image for subsequent tracking and analysis. Extract the optimal dough state features from the database. These features should include indicators such as shape, texture, and uniformity. This will serve as the benchmark for evaluation. Ensure that the feature data is complete and accurate to ensure the validity of the evaluation results. Perform feature extraction on the final kneading image. Use the same image processing method as before to extract key features such as shape, area, and contour. Use OpenCV's findContours and cv2.ContourArea and other functions are used to obtain the shape features of the image. According to the defined evaluation criteria (such as threshold-based judgment or machine learning model), the state evaluation value of the final image is calculated. The evaluation values ​​of each feature can be combined using a simple weighted average method to form a comprehensive state evaluation value. The calculated final dough state evaluation value is stored in the database, and its corresponding timestamp, image and related parameters are recorded for subsequent analysis and reference. The historical kneading parameters, state evaluation values ​​and final images are extracted from the database for performance analysis and model training. The integrity of the data set is ensured, including different states and corresponding evaluation values ​​in multiple kneading processes. A suitable self-reinforcement learning algorithm is selected, such as Q-learning and deep Q network (DQN). Or policy gradient methods, these algorithms can continuously optimize the decision-making process through feedback, determine the state space, action space and reward mechanism to guide the learning process. For example, the state space can be the characteristics of the dough, and the action space can be the adjustment of the kneading parameters. The extracted data is used to train the intelligent kneading control model. Through continuous iteration and adjustment, the model will learn the relationship between the optimal kneading parameters and the dough state evaluation value. During the training process, the model performance is monitored in real time, and the learning rate and other hyperparameters are adjusted to ensure that the model can converge effectively. After the model training is completed, a verification test is performed to evaluate the performance of the model in real-time kneading control to ensure that it can effectively improve the dough quality. The generated intelligent kneading control model is deployed to actual operations, and the parameters are adjusted in real time to further optimize the kneading process. .

[0074] In this embodiment, a system for selecting an intelligent dough kneading mode is provided, which is used to execute the method for selecting an intelligent dough kneading mode as described above, including: An image reconstruction module is used to obtain a real-time dough monitoring image in a dough kneading machine; perform edge sharpening and reconstruction on the real-time dough monitoring image, and perform time series frame sequence fitting to construct a dough monitoring time series frame sequence; The timing state diagram module is used to identify the dough contour frame by frame in the dough monitoring timing frame sequence, and to conduct in-depth mining of the dough timing state to construct a dough timing state diagram; The activity trend prediction module is used to mine the dynamic changes of yeast particles frame by frame and predict the activity trend at multiple moments in the dough monitoring time series, so as to extract the optimal kneading time point; A dynamic dough kneading module is used to match the time state of the dough timing state diagram based on the optimal dough kneading time point, and then make dynamic dough kneading parameter decisions to generate dynamic dough kneading parameters; Parameter fine-tuning module, which is used to evaluate dough state based on dynamic kneading parameters, and to fine-tune and optimize instant kneading parameters, and to build an instant kneading fine-tuning strategy; Intelligent dough kneading control is used to perform comprehensive state evaluation based on the instant dough kneading fine-tuning strategy and self-reinforcement dough kneading performance learning to generate an intelligent dough kneading control model.

[0075] The present invention uses sensors such as cameras to obtain images of dough in a dough kneading machine in real time, provides raw data for subsequent analysis, performs edge sharpening processing on the real-time images obtained, highlights the contours and details of the dough, and ensures that subsequent image analysis is more accurate. Multiple frames of images are fitted in time-series frame order to construct a time-series frame sequence for dough monitoring, which helps the system identify dynamic changes of the dough during the kneading process, tracks contour changes of the dough in real time, and ensures that the system can detect slight changes in the dough morphology. After constructing a dough time-series state diagram, the change trend of the dough at different time points can be understood, providing accurate data for further analysis and decision-making. In-depth mining of the time-series state helps to identify potential kneading problems (such as over-kneading or over-relaxation), and improves the system's ability to identify and predict the dough state. By predicting yeast activity at multiple moments, the system can dynamically adjust parameters during the kneading process to ensure the optimal fermentation state of the yeast, accurately calculate the optimal kneading time point, and avoid kneading the dough when the yeast activity is not appropriate, thereby improving the fermentation effect and dough quality. The activity trend prediction module improves the control accuracy during the dough fermentation process. , ensure that the dough can reach the ideal fermentation state, ensure the accuracy of the kneading time by matching the optimal kneading time point, avoid over-kneading or under-kneading, dynamically adjust the kneading parameters according to the real-time feedback of the dough state, make the kneading process more personalized, and adapt to the needs of different doughs, accurate kneading control can improve the kneading efficiency and ensure that each stage of the dough is in the best state, by fine-tuning the kneading parameters, the system can respond to the changes in the dough state in time to ensure that the dough is always in the best state, no matter what changes occur in the dough during the kneading process, the system can flexibly adjust the kneading parameters to reduce the impact caused by environmental changes or other factors, the fine-tuning strategy can help maintain the consistency of the quality of each kneading and avoid quality fluctuations caused by operational errors, through self-reinforcement learning, the system can continuously accumulate experience and data, so as to continuously optimize the kneading control strategy, the generated intelligent kneading control model can automatically adapt to various production environments, improve the intelligence of the kneading process, and reduce manual intervention, with the accumulation of more data, the intelligent model will become more and more accurate, and continuously improve the kneading efficiency and product quality consistency.

[0076] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0077] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for selecting an intelligent dough kneading mode, characterized in that: The following steps are involved: Step S1: obtaining a real-time dough monitoring image in a dough kneading machine; performing edge sharpening and reconstruction on the real-time dough monitoring image, and performing time series frame sequence fitting to construct a dough monitoring time series frame sequence; Step S2: Recognize the dough contour frame by frame on the dough monitoring time sequence frame sequence, and conduct deep mining of the dough time sequence state to construct a dough time sequence state diagram; Step S3: mining the dynamic changes of yeast particles frame by frame and predicting the activity trend at multiple moments in the dough monitoring time series frame sequence, so as to extract the optimal kneading time point; Step S4: performing time state matching on the dough timing state diagram based on the optimal dough kneading time point, and then making dynamic dough kneading parameter decisions to generate dynamic dough kneading parameters; Step S5: evaluating the dough state based on the dynamic kneading parameters, and fine-tuning and optimizing the instant kneading parameters to construct an instant kneading fine-tuning strategy; Step S6: Perform comprehensive state evaluation based on the instant dough kneading fine-tuning strategy, and conduct self-reinforcement dough kneading performance learning to generate an intelligent dough kneading control model.

2. The method for selecting an intelligent dough kneading mode according to claim 1, characterized in that: The specific steps of step S1 are: Step S11: obtaining a real-time dough monitoring image in a dough kneading machine; Step S12: performing edge sharpening and reconstruction on the real-time dough monitoring image to construct a detail-optimized and reconstructed dough image; Step S13: extracting frame-by-frame image changes of the detail-optimized and reconstructed dough image to obtain a dough monitoring image of each frame; Step S14: Calculate the time stamp of sequential frame acquisition for the detail optimized and reconstructed dough image to obtain the acquisition time stamp of each frame; Step S15: performing time-series frame sequence fitting on each frame of the dough monitoring image according to the acquisition timestamp of each frame, and constructing a dough monitoring time-series frame sequence.

3. The method for selecting an intelligent dough kneading mode according to claim 2, characterized in that: The specific steps of step S12 are: Perform pixel-by-pixel recognition on the real-time dough monitoring image to extract all pixel points in the image; Perform global pixel average calculation on all pixels in the image to generate the global pixel average of the image; Image noise suppression is performed according to the global pixel average value of the image, thereby obtaining a denoised and optimized dough monitoring image; Perform histogram equalization on the denoised and optimized dough monitoring image to generate an image grayscale histogram; Perform grayscale distribution stretching on the image grayscale histogram and calculate the grayscale value of the stretched pixel; performing contrast enhancement according to the grayscale values ​​of the stretched pixels to obtain a contrast enhanced dough monitoring image; Perform edge gradient information analysis on the texture detail enhanced dough monitoring image to obtain an edge map; Performing slight texture change sharpening on the edge map to generate a slight texture change sharpened edge map; The contrast enhanced dough monitoring image is reconstructed by edge sharpening according to the micro-texture change sharp edge map to construct a detail optimized reconstructed dough image.

4. The method for selecting an intelligent dough kneading mode according to claim 1, characterized in that: The specific steps of step S2 are: Step S21: performing dough contour recognition frame by frame on the dough monitoring time series frame sequence, and extracting the dough contour line of each frame; Step S22: Tracking the dynamic changes of the dough contour line of each frame to obtain a dynamic change trajectory of the dough contour; Step S23: performing contour change trend analysis on the dynamic change trajectory of the dough contour to generate dough contour change trend characteristics; Step S24: evaluating the dough surface roughness of the dough monitoring time series frame sequence to generate a dough surface roughness time series curve; Step S25: performing frame-by-frame analysis of the dough crack generation situation on the dough monitoring time series frame sequence to obtain crack evolution characteristics in the dough; Step S26: Conduct in-depth exploration of the dough time-series state based on the dough surface roughness time-series curve, dough contour change trend characteristics, and dough crack evolution characteristics to construct a dough time-series state diagram.

5. The method for selecting an intelligent dough kneading mode according to claim 1, characterized in that: The specific steps of step S3 are: Step S31: mining the dynamic changes of yeast particles frame by frame in the dough monitoring time series frame sequence to extract the dynamic change characteristics of yeast particles; Step S32: performing yeast temporal activity analysis according to the dynamic change characteristics of yeast particles to generate a yeast temporal activity curve; Step S33: Evolving the activity variation law of the yeast time series activity curve to obtain the yeast activity variation law; Step S34: performing multi-time activity trend prediction based on the yeast activity change rule, thereby generating yeast activity trend prediction data at multiple times; Step S35: Calculate the optimal kneading time based on the yeast activity trend prediction data at multiple moments, so as to extract the optimal kneading time point.

6. The method for selecting an intelligent dough kneading mode according to claim 5, characterized in that: The specific steps of step S35 are: Perform multi-time activity trend peak calculation on yeast activity trend prediction data at multiple times, and extract the activity trend peak of each time period; The activity trend peak in each time period is used to identify the inflection point of activity change and marked as the highest peak point of yeast activity; Define preset kneading time parameters; Calculate the kneading buffer time at the highest peak point of yeast activity according to the preset kneading time parameters to obtain the kneading buffer time at the highest peak point; The optimal kneading time is calculated based on the kneading buffer time at the highest peak point to obtain the optimal kneading time point.

7. The method for selecting an intelligent dough kneading mode according to claim 1, characterized in that: The specific steps of step S4 are: Step S41: performing time state matching on the dough time sequence state diagram based on the optimal kneading time point, and extracting the dough time sequence state characteristics at the optimal kneading time point; Step S42: Analyze the optimal dough state of the dough time sequence state diagram to generate optimal dough state characteristics; Step S43: performing an optimal state deviation analysis on the optimal dough state characteristics based on the dough temporal state characteristics at the optimal dough kneading time point to obtain the dough state deviation characteristics at the optimal dough kneading time point; Step S44: making dynamic kneading parameter decisions based on the dough state deviation characteristics at the optimal kneading time point to generate dynamic kneading parameters.

8. The method for selecting an intelligent dough kneading mode according to claim 1, characterized in that: The specific steps of step S5 are: Step S51: performing real-time dough kneading control based on dynamic dough kneading parameters, and acquiring real-time dough kneading monitoring images; Step S52: performing dough shape analysis on the real-time dough kneading monitoring image to obtain the real-time dough kneading shape; Step S53: performing dough state evaluation on the real-time dough shape during kneading to generate a real-time dough state evaluation value; Step S54: Based on the real-time kneading state evaluation value, the dynamic kneading parameters are fine-tuned and optimized to construct an instant kneading fine-tuning strategy.

9. The method for selecting an intelligent dough kneading mode according to claim 1, characterized in that: The specific steps of step S6 are: Step S61: extracting a final dough kneading image according to the real-time dough kneading monitoring image; Step S62: performing a comprehensive state evaluation on the final dough kneading image according to the optimal dough state characteristics to obtain a final dough state evaluation value; Step S63: Based on the final dough state evaluation value, the instant dough kneading fine-tuning strategy is self-reinforced to learn the dough kneading performance, thereby generating an intelligent dough kneading control model.

10. An intelligent dough kneading mode selection system, characterized in that: The method for selecting the intelligent dough kneading mode according to claim 1 comprises: An image reconstruction module is used to obtain a real-time dough monitoring image in a dough kneading machine; perform edge sharpening and reconstruction on the real-time dough monitoring image, and perform time series frame sequence fitting to construct a dough monitoring time series frame sequence; The timing state diagram module is used to identify the dough contour frame by frame in the dough monitoring timing frame sequence, and to conduct in-depth mining of the dough timing state to construct a dough timing state diagram; The activity trend prediction module is used to mine the dynamic changes of yeast particles frame by frame and predict the activity trend at multiple moments in the dough monitoring time series, so as to extract the optimal kneading time point; A dynamic dough kneading module is used to match the time state of the dough timing state diagram based on the optimal dough kneading time point, and then make dynamic dough kneading parameter decisions to generate dynamic dough kneading parameters; Parameter fine-tuning module, which is used to evaluate dough state based on dynamic kneading parameters, and to fine-tune and optimize instant kneading parameters, and to build an instant kneading fine-tuning strategy; Intelligent dough kneading control is used to perform comprehensive state evaluation based on the instant dough kneading fine-tuning strategy and self-reinforcement dough kneading performance learning to generate an intelligent dough kneading control model.

Images