A machine vision-based intelligent control system for tea rolling

Abstract

The present application relates to the field of tea rolling technology, and specifically discloses a tea rolling intelligent control system based on machine vision, which comprises the following steps: extracting tea state data in the rolling process, obtaining current rolling parameters, constructing a dynamic data set, obtaining rolling parameters under different tea states, establishing a tea rolling state database, inputting the dynamic data set into the tea rolling state database for adaptability analysis, analyzing the difference between the current rolling state and the optimal working condition, judging whether the rolling parameters need to be adjusted, confirming the adjustment direction if necessary, performing adaptive rolling parameter adjustment, predicting the tea state change trend according to the dynamic data set, adjusting the rolling parameter dynamic adaptation to the tea state change, comparing the real-time environmental temperature and humidity fluctuation information with the dynamic data set and the historical optimal working condition characteristics, screening the rolling parameter adaptation rules of different types of tea, and generating a personalized fine-tuning scheme, which is conducive to improving the stability of tea rolling quality.

Description

Technical Field

[0001] This invention relates to the field of tea rolling technology, specifically to an intelligent control system for tea rolling based on machine vision. Background Technology

[0002] In tea processing, rolling is a core step in shaping the appearance and improving the internal quality of tea leaves. Precise control of its process parameters directly determines the quality of the finished tea. Currently, tea rolling production in the industry still heavily relies on manual experience. The core process curves, such as pressure, time, and rotation speed, are entirely set by the rolling master based on subjective experience, considering the tea variety, the age of the fresh leaves, and the state of the withered leaves. Different masters have different experience systems, and even within the same batch of tea, parameter deviations can occur due to fluctuations in the master's condition, leading to instability in the rolling effect and potential risks to the quality of the finished tea.

[0003] In existing technologies, the current rolling production mode lacks a real-time process feedback and adaptive adjustment mechanism. During the rolling process, it is impossible to dynamically perceive changes in the state of tea leaves such as shape, tightness, and juice overflow. The processing effect can only be judged by sensory evaluation afterward. During the rolling process, it is impossible to identify whether the current rolling parameters are suitable. Deviations in rolling parameters are difficult to correct in time, resulting in huge fluctuations in tea quality. Key indicators such as the regularity and tightness of the finished tea leaves are inconsistent, making it difficult to meet the standardization requirements of large-scale production.

[0004] Therefore, the present invention provides an intelligent control system for tea kneading based on machine vision. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent control system for tea kneading based on machine vision, so as to solve the above-mentioned background problems.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A machine vision-based intelligent control system for tea leaf kneading includes:

[0008] Tea leaf state extraction module: Extracts tea leaf state data during the rolling process, obtains current rolling parameters, and constructs a dynamic dataset;

[0009] Adaptation Analysis and Adjustment Module: Obtain rolling parameters under different tea conditions, establish a tea rolling state database, input dynamic datasets into the tea rolling state database for adaptation analysis, quantify and evaluate the difference between the current rolling state and the optimal working condition through feature deviation, determine whether the rolling parameters need to be adjusted, and if so, confirm the adjustment direction;

[0010] Adaptive adjustment module: Based on the adjustment direction of the kneading parameters, it performs adaptive kneading parameter adjustment, and predicts the trend of tea state change based on dynamic dataset, and dynamically adjusts the kneading parameters to adapt to the changes in tea state.

[0011] Comparison and screening module: Obtain tea characteristic data after adjusting the rolling parameters, combine it with real-time environmental temperature and humidity fluctuation information, compare it with dynamic datasets and historical best working condition characteristics, screen the rolling parameter adaptation rules of different types of tea, and generate personalized fine-tuning schemes.

[0012] Furthermore, the tea condition data includes the tea leaf formation rate, tightness, and leaf juice overflow:

[0013] Obtain the current tea leaf state image, extract data from the tea leaf state image, and obtain tea leaf state data;

[0014] Image processing algorithms are used to extract tea state data from the obtained tea state images, and a dynamic dataset is constructed.

[0015] Furthermore, construct a dynamic dataset:

[0016] Extract the tea leaf contour from the tea leaf condition image, obtain the tea leaf region for each tea leaf, calculate the aspect ratio and curvature based on the tea leaf region, and perform standardization processing.

[0017] The aspect ratio and curvature are processed to obtain the strip formation coefficient of each tea leaf. The strip formation coefficients of all tea leaves in the tea leaf state image are averaged to obtain the strip regularity coefficient.

[0018] Based on images of tea leaves, the contour compactness and surface texture uniformity of tea leaves are calculated.

[0019] The compactness coefficient of each tea leaf is obtained by multiplying the contour compactness and the surface texture uniformity. The compactness coefficient of all tea leaves in the tea leaf state image is then averaged to obtain the compactness coefficient.

[0020] Based on the tea leaf regions already acquired in the tea leaf condition image, the color saturation difference and surface highlight ratio of the tea leaves are calculated and then standardized.

[0021] The color saturation difference and the surface highlight ratio are summed to obtain the juice overflow coefficient of each tea leaf. The juice overflow coefficients of all tea leaves in the tea leaf state image are averaged to obtain the leaf juice overflow degree coefficient.

[0022] Obtain the current kneading parameters, including kneading speed and kneading pressure;

[0023] By integrating the regularity coefficient of the strip, the compactness coefficient, the sap overflow coefficient, and the rolling parameters according to the same timestamp, a dynamic dataset is obtained.

[0024] Furthermore, establish a database of tea leaf rolling conditions:

[0025] Obtain appropriate rolling parameters for different tea conditions, and archive the data in layers according to two dimensions: tea type and rolling stage;

[0026] First, by classifying tea varieties, the suitable rolling parameters for tea of ​​the same variety are summarized. Then, the rolling stages are divided according to the suitable rolling parameters for tea of ​​the same variety. Based on the different rolling stages of tea during the rolling process, the suitable rolling parameters for different tea states are matched one by one, and a tea rolling state database is constructed.

[0027] Further, an adaptation analysis was conducted:

[0028] Input the tea state data from the dynamic dataset into the tea rolling state database, and output the corresponding adaptive rolling parameters.

[0029] The adaptive kneading parameters and the kneading parameters are quantified separately to obtain the adaptive kneading parameter data and the kneading parameter data.

[0030] The difference between the adapted kneading parameter data and the kneading parameter data is processed to obtain the parameter deviation value;

[0031] If the parameter deviation value is greater than or equal to the upper limit of the deviation range, or less than or equal to the lower limit of the deviation range, then the current processing parameters are not suitable for the current tea rolling and need to be adjusted.

[0032] Furthermore, confirm the direction of the adjustment:

[0033] If the parameter deviation value is greater than or equal to the upper limit of the deviation range, the current kneading parameter exceeds the suitable kneading parameter under the same product category and kneading stage, and the current kneading parameter needs to be adjusted downward.

[0034] If the parameter deviation value is less than or equal to the lower limit of the deviation range, the current kneading parameter is lower than the suitable kneading parameter for the same product category and kneading stage, and the current kneading parameter needs to be adjusted upward.

[0035] Furthermore, adaptive kneading parameters are adjusted:

[0036] Based on the adjustment direction and parameter deviation value of the kneading parameters, the initial adjustment of the kneading parameters is carried out, and the initial adjustment amount is the parameter deviation value;

[0037] When the adjustment direction is downward, that is, when the parameter deviation value is greater than or equal to the upper limit of the deviation range, the corresponding kneading parameter is adjusted downward according to the initial adjustment.

[0038] When the adjustment direction is upward, that is, when the parameter deviation value is less than or equal to the lower limit of the deviation range, the corresponding kneading parameter is adjusted upward according to the initial adjustment.

[0039] Furthermore, predict the trend of tea leaf condition changes based on dynamic datasets:

[0040] Extract tea state data with consecutive timestamps from dynamic datasets, calculate the slope of change of the regularity coefficient of strips, the compactness coefficient, and the degree of sap overflow coefficient within the same time period, and construct a tea state change trend prediction model through linear fitting.

[0041] Based on the tea state change trend prediction model, after initial adjustment, the tea state data within the time period after initial adjustment is obtained. The tea state data after initial adjustment is input into the tea state change trend prediction model, and the predicted values ​​of the change direction and change magnitude of each state coefficient of tea in the next time period are output to predict the tea state change trend.

[0042] Furthermore, the kneading parameters are dynamically adjusted to adapt to changes in the tea's state:

[0043] Based on the predicted trend of tea state changes, the kneading parameters after the initial adjustment are dynamically adapted to the changes in tea state.

[0044] If the predicted regularity coefficient of the tea leaves shows a continuous upward trend, and the current regularity coefficient of the tea leaves is close to the threshold of the optimal range for the same category and stage, then reduce the adjustment range of the rolling speed to avoid excessive strip formation of tea leaves leading to an increase in the broken leaf rate.

[0045] If the predicted rate of increase in the compactness coefficient is slowing down, and the current compactness coefficient is below the optimal range threshold, then the rolling pressure adjustment range should be increased to enhance the compactness of the tea leaves.

[0046] Furthermore, comparisons were made with dynamic datasets and historical best operating condition features:

[0047] Acquire adaptively adjusted tea state data and real-time environmental temperature and humidity fluctuation information during the rolling process;

[0048] The adaptively adjusted tea characteristic data and real-time environmental temperature and humidity fluctuation information are compared and analyzed in multiple dimensions with the historical best working condition characteristic data of the same category and the same kneading stage in the tea kneading state database.

[0049] Construct a multi-dimensional comparison and evaluation index system, which includes feature similarity, working condition matching degree, and environmental adaptability;

[0050] The feature similarity, working condition matching degree, and environmental adaptability are summed to obtain the comprehensive matching coefficient;

[0051] If the overall matching coefficient is greater than or equal to the overall matching threshold, it means that the adjusted kneading parameters and the corresponding environment are suitable for the current kneading stage of the corresponding tea category. The kneading parameters, environmental temperature and humidity conditions, and tea characteristic data are archived according to the category-stage-environment dimension to obtain the kneading parameter adaptation rules.

[0052] The beneficial effects of this invention are as follows:

[0053] 1. This invention uses industrial vision sensors to collect real-time images of tea leaves during the rolling process, extracting core data such as strip formation rate, tightness, and sap overflow. These data are quantified into strip regularity coefficients, tightness coefficients, and sap overflow coefficients. A dynamic dataset is constructed by combining this data with the current rolling speed and pressure parameters. Simultaneously, based on the practical experience of rolling masters, a tea rolling state database is established, categorized by tea type and rolling stage. The dynamic dataset is compared with the optimal parameters in the database to quantitatively assess the deviation between the current rolling state and the optimal working condition, and to clarify the direction of parameter adjustment. This solves the problem of traditional rolling relying on manual experience and lacking objective state assessment and quantitative basis, providing a precise time-related data foundation for subsequent adaptive control and ensuring the targeted and scientific nature of parameter adjustments.

[0054] 2. Based on the parameter adjustment direction, this invention adopts a two-layer control strategy of initial adjustment + dynamic fine-tuning, and combines a tea state change trend prediction model to achieve adaptive optimization of rolling parameters. At the same time, it integrates parameter adjustment effects, real-time environmental temperature and humidity fluctuation information and historical optimal working condition data to screen the rolling parameter adaptation rules of different types of tea in specific temperature and humidity ranges, and generates a personalized fine-tuning scheme that includes step-by-step parameter adjustment and temperature and humidity linkage compensation, thereby improving the stability of tea rolling quality. Attached Figure Description

[0055] The invention will now be further described with reference to the accompanying drawings.

[0056] Figure 1 This is a flowchart of a machine vision-based intelligent control system for tea kneading, as described in an embodiment of the present invention.

[0057] Figure 2 This is a flowchart illustrating a machine vision-based intelligent control system for tea kneading, as described in an embodiment of the present invention. Detailed Implementation

[0058] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0059] Example 1

[0060] Please see Figure 1 As shown in the embodiment of the present invention, an intelligent control system for tea kneading based on machine vision is proposed. This invention primarily addresses the problem that the tea kneading process heavily relies on manual experience, with the pressure-speed curve entirely set by the kneading master based on the tea variety, maturity, and state of the withered leaves, lacking process feedback and adaptive adjustment. This results in large fluctuations in tea quality and difficulty in standardization. The invention uses machine vision technology to collect and analyze tea kneading state data in real time, constructing a multi-dimensional tea kneading state database. Combined with dynamic adaptation analysis and trend prediction, it achieves adaptive control of kneading parameters, ultimately refining process adaptation rules and generating personalized fine-tuning schemes. This solves the industry pain points of reliance on manual experience and lack of dynamic feedback, achieving standardization and precision in the tea kneading process. Specifically, it includes the following modules:

[0061] Tea leaf state extraction module: Extracts tea leaf state data during the rolling process, obtains current rolling parameters, and constructs a dynamic dataset;

[0062] Specifically, during the kneading process, industrial vision sensors are used to collect real-time images of the tea leaves to obtain images of the current state of the tea leaves. Data is then extracted from these images to obtain tea leaf state data.

[0063] The industrial vision sensor includes an industrial-grade high-speed color camera and an image processing unit. The industrial-grade high-speed color camera is responsible for capturing images of the morphological changes of tea leaves during the rolling process, while the image processing unit performs preprocessing on the acquired images, including noise reduction and enhancement, to improve image quality.

[0064] The obtained tea state images are processed using image processing algorithms to extract tea state data and construct a dynamic dataset. The tea state data includes the tea leaf formation rate, tightness, and sap overflow.

[0065] Extract the tea leaf contour from the tea leaf condition image, obtain the tea leaf region for each tea leaf, calculate the aspect ratio and curvature based on the tea leaf region, and perform standardization processing.

[0066] It should be noted that the aspect ratio is obtained by comparing the length of the longest axis with the length of the shortest axis, and the curvature is obtained by comparing the actual profile length with the length of the diagonal of the smallest circumscribed rectangle.

[0067] The aspect ratio and curvature are processed to obtain the strip formation coefficient of each tea leaf. The strip formation coefficients of all tea leaves in the tea leaf state image are averaged to obtain the strip regularity coefficient.

[0068] Based on images of tea leaves, the contour compactness and surface texture uniformity of tea leaves are calculated.

[0069] It should be noted that the contour compactness is obtained by processing the ratio of the tea leaf contour area to the area of ​​the smallest circumscribed circle, and the surface texture uniformity is obtained by calculating the texture entropy value using the gray-level co-occurrence matrix and then normalizing it.

[0070] The compactness coefficient of each tea leaf is obtained by multiplying the contour compactness and the surface texture uniformity. The compactness coefficient of all tea leaves in the tea leaf state image is then averaged to obtain the compactness coefficient.

[0071] Based on the tea leaf regions already acquired in the tea leaf condition image, the color saturation difference and surface highlight ratio of the tea leaves are calculated and then standardized.

[0072] It should be noted that the color saturation difference is obtained by subtracting the saturation value of the current tea area in the HSV color space from the saturation reference value of the tea in the initial stage of rolling. The surface highlight ratio is obtained by subtracting the number of pixels in the tea area whose brightness value is higher than a set threshold from the total number of pixels in the tea area.

[0073] The color saturation difference and the surface highlight ratio are summed to obtain the juice overflow coefficient of each tea leaf. The juice overflow coefficients of all tea leaves in the tea leaf state image are averaged to obtain the leaf juice overflow degree coefficient.

[0074] Obtain the current kneading parameters, including kneading speed and kneading pressure;

[0075] The regularity coefficient of strips, the compactness coefficient, the sap overflow coefficient, and the rolling parameters are integrated according to the same time stamp to obtain a dynamic dataset;

[0076] It should be noted that the purpose of this module is to collect core data on the state of tea leaves during the rolling process in real time (coefficient of regularity of tea leaves, coefficient of compactness of tea leaves, coefficient of degree of sap overflow) and key rolling parameters (rolling speed, rolling pressure), and to integrate the data according to a unified timestamp to build a dynamic dataset. This provides a real-time, accurate and time-correlated data foundation for subsequent modules to conduct difference assessments between the current rolling state and the optimal working conditions and to adjust parameters adaptively. This ensures that subsequent control decisions are based on actual rolling working condition data. At the same time, it provides continuous process data for the final screening of tea rolling parameter adaptation patterns and the generation of personalized fine-tuning schemes, ensuring the feasibility and accuracy of the entire rolling adaptive control process.

[0077] Adaptation Analysis and Adjustment Module: Obtains rolling parameters under different tea leaf conditions, establishes a tea leaf rolling state database, inputs dynamic datasets into the tea leaf rolling state database for adaptation analysis, quantifies and evaluates the difference between the current rolling state and the optimal working condition through feature deviation, determines whether the rolling parameters need to be adjusted, and if so, then...

[0078] Specifically, based on the experience of the kneading masters, appropriate kneading parameters are obtained for different tea conditions, including appropriate kneading parameters for different types of tea, different kneading stages, and different tea conditions. Among them, the tea condition data includes the regularity coefficient of the strips, the compactness coefficient, and the degree of leaf juice overflow coefficient, and the appropriate kneading parameters include the kneading speed and the kneading pressure.

[0079] Based on the appropriate rolling parameters for different tea conditions, the data is stratified and archived according to two dimensions: tea category and rolling stage.

[0080] First, by classifying tea varieties, the suitable rolling parameters for tea of ​​the same variety are summarized. Then, the rolling stages are divided according to the suitable rolling parameters for tea of ​​the same variety. Based on the different rolling stages of tea during the rolling process, the suitable rolling parameters for different tea states are matched one by one to construct a tea rolling state database.

[0081] It should be noted that the purpose of building the tea rolling state database is to obtain the appropriate rolling parameters for the current state by inputting the tea category and the corresponding rolling stage tea state data into the tea rolling state database.

[0082] Based on the tea type and rolling stage, the tea rolling state database is used to locate the tea rolling state data. The tea state data in the dynamic dataset is input into the tea rolling state database, and the corresponding adaptive rolling parameters are output.

[0083] The compatibility analysis of the obtained adaptive kneading parameters with the kneading parameters of the current kneading stage is performed.

[0084] The adaptive kneading parameters and the kneading parameters are quantified separately to obtain the adaptive kneading parameter data and the kneading parameter data.

[0085] The difference between the adaptive kneading parameter data and the kneading parameter data is processed to obtain the parameter deviation value, and the parameter deviation value is compared with the deviation range threshold.

[0086] If the parameter deviation value is within the deviation range threshold, it means that the current kneading parameters are suitable for the current tea kneading.

[0087] If the parameter deviation value is greater than or equal to the upper limit of the deviation range, or less than or equal to the lower limit of the deviation range, it means that the current processing parameters are not suitable for the current tea rolling and need to be adjusted.

[0088] It should be noted that the upper and lower limits of the deviation range refer to the reasonable range of fluctuation allowed for the kneading parameters (kneading speed and kneading pressure) set based on the quality requirements of different tea varieties, the process characteristics of the kneading stage, and the combination of historical best production data and the practical experience of kneading masters. These parameters are used to determine whether the current kneading parameters deviate from the quantitative parameter standards for the suitable working conditions.

[0089] If the parameter deviation value is greater than or equal to the upper limit of the deviation range, it means that the current kneading parameter exceeds the suitable kneading parameter under the same product category and kneading stage, and the current kneading parameter needs to be adjusted downward.

[0090] If the parameter deviation value is less than or equal to the lower limit of the deviation range, it means that the current kneading parameters are lower than the suitable kneading parameters for the same product category and kneading stage, and the current kneading parameters need to be adjusted upward.

[0091] It should be noted that the purpose of this module is to build a database of tea rolling state, compare the dynamic dataset with the optimal adaptation parameters in the tea rolling state database, use the parameter deviation value as the core quantitative indicator, obtain the degree of difference between the current rolling parameters and the optimal working conditions, and clarify the direction of parameter adjustment. At the same time, the database's hierarchical archiving structure according to tea category and rolling stage can ensure the accuracy and relevance of the adaptation analysis, and avoid the problem of mixing parameters for different tea categories and different rolling stages.

[0092] Example 2

[0093] like Figure 2 As shown in the embodiment of the present invention, a machine vision-based intelligent control system for tea leaf kneading includes the following modules:

[0094] Adaptive adjustment module: Based on the adjustment direction of the kneading parameters, it performs adaptive kneading parameter adjustment, and predicts the trend of tea state change based on dynamic dataset, and dynamically adjusts the kneading parameters to adapt to the changes in tea state.

[0095] Specifically, based on the direction of the kneading parameter adjustment (upward or downward) and the parameter deviation value, the initial adjustment of the kneading parameters is carried out, and the initial adjustment amount is the parameter deviation value.

[0096] It should be noted that the initial adjustment of the rolling parameters is positively correlated with the parameter deviation value. That is, the larger the parameter deviation value, the higher the initial adjustment. At the same time, the initial adjustment should refer to the historical parameter adjustment experience thresholds of the same type and rolling stage in the tea rolling status database to avoid sudden changes in rolling parameters.

[0097] When the adjustment direction is downward, that is, when the parameter deviation value is greater than or equal to the upper limit of the deviation range, the corresponding kneading parameter is adjusted downward according to the initial adjustment.

[0098] When the adjustment direction is upward, that is, when the parameter deviation value is less than or equal to the lower limit of the deviation range, the corresponding kneading parameter is adjusted upward according to the initial adjustment.

[0099] Extract tea state data (strip regularity coefficient, compactness coefficient, and sap overflow coefficient) from the dynamic dataset with continuous timestamps, calculate the slope of each coefficient change within the same time period, and construct a tea state change trend prediction model through linear fitting.

[0100] It should be noted that the set time period refers to a dynamic dataset segment containing continuous and equal time interval timestamps selected for analyzing the continuous change pattern of tea state coefficients. The length of the time period can be set according to the characteristics of the tea category and the process requirements of the rolling stage to ensure that it can cover the minimum time period in which the state of tea fluctuates significantly. The slope of change refers to the slope of the fitted line calculated by linear fitting within the set time period, with the timestamp as the horizontal axis and the state coefficient of a single type of tea as the vertical axis. The sign of the slope indicates the direction of change of the state coefficient of that type of tea, and the magnitude of the absolute value of the slope indicates the rate of change of the state coefficient of that type of tea.

[0101] Based on the tea state change trend prediction model, after the initial adjustment, the tea state data within the time period after the initial adjustment is obtained. The tea state data after the initial adjustment is input into the tea state change trend prediction model, and the predicted values ​​of the change direction and change magnitude of each state coefficient of tea in the next time period are output to predict the tea state change trend.

[0102] Based on the predicted trend of tea state changes, the kneading parameters after the initial adjustment are dynamically adapted to the changes in tea state.

[0103] For example, if the predicted strip regularity coefficient shows a continuous upward trend, and the current strip regularity coefficient is close to the optimal range threshold of the same category and stage, then reduce the adjustment range of the rolling speed to avoid excessive strip formation of tea leaves leading to an increase in the broken leaf rate.

[0104] If the growth rate of the compressibility coefficient is predicted to slow down, and the current compressibility coefficient is below the optimal range threshold, then the adjustment range of the rolling pressure should be increased to enhance the compressibility of the tea leaves.

[0105] It should be noted that the optimal range threshold refers to a reasonable range of standards set for core state data such as the regularity coefficient of tea strips, the compactness coefficient, and the degree of leaf juice overflow coefficient under the same stage of the same type of tea in the tea rolling state database, which is based on historical best production data, the practical experience of rolling masters, and the quality test results of finished tea. It is used to judge whether the current tea rolling state is close to the optimal working condition.

[0106] It should be noted that the function of this module is to achieve adaptive optimization of the rolling parameters through a two-layer parameter control strategy of initial adjustment + dynamic fine-tuning, combined with the prediction of the tea state change trend. This solves the problem of tea state being easily out of control under the traditional fixed parameter rolling mode, and ensures that the tea always approaches the optimal working condition throughout the entire rolling cycle.

[0107] Comparison and screening module: Obtain tea characteristic data after adjusting the rolling parameters, combine it with real-time environmental temperature and humidity fluctuation information, compare it with dynamic datasets and historical best working condition characteristics, screen the rolling parameter adaptation rules of different types of tea, and generate personalized fine-tuning schemes.

[0108] Specifically, after the kneading parameters are adaptively adjusted, the tea leaves are captured in real time using an industrial vision sensor to obtain the adaptively adjusted tea leaf state data, including the adaptively adjusted strip regularity coefficient, compactness coefficient, and leaf juice overflow coefficient.

[0109] High-precision temperature and humidity sensors are used to collect real-time environmental temperature and humidity fluctuation information during the kneading process (real-time humidity value and the range of temperature and humidity changes within a set time period).

[0110] The adaptively adjusted tea characteristic data and real-time environmental temperature and humidity fluctuation information are compared and analyzed in multiple dimensions with the historical best working condition characteristic data of the same category and the same kneading stage in the tea kneading state database.

[0111] Construct a multi-dimensional comparison and evaluation index system, which includes feature similarity, working condition matching degree, and environmental adaptability;

[0112] It should be noted that feature similarity is obtained by calculating the Euclidean distance between the adjusted tea feature data and the corresponding coefficients of the historical best working condition features, and then normalizing the result. The smaller the Euclidean distance, the higher the feature similarity. Working condition matching degree is obtained by comparing the overlap between the adjusted rolling parameters and the historical best working condition parameters, and then weighting the result by combining the change range of the tea state coefficient before and after parameter adjustment. Environmental adaptability is obtained by analyzing the deviation values ​​between real-time temperature and humidity and standard temperature and humidity, and then quantitatively evaluating the influence weight of the deviation value on the rolling state of the tea.

[0113] The feature similarity, working condition matching degree, and environmental adaptability are summed to obtain the comprehensive matching coefficient;

[0114] If the overall matching coefficient is greater than or equal to the overall matching threshold, it means that the currently adjusted kneading parameters and the corresponding environment are suitable for the current kneading stage of the corresponding tea category. The kneading parameters, environmental temperature and humidity conditions, and tea characteristic data are archived according to the category-stage-environment dimension to obtain the kneading parameter adaptation rules.

[0115] If the overall matching coefficient is less than the overall matching threshold, it means that it is not applicable to the current tea rolling stage;

[0116] Based on the adaptation rules of the kneading parameters obtained through screening, a personalized fine-tuning scheme is generated, which includes a basic parameter scheme and a dynamic correction scheme.

[0117] The basic parameter scheme is the initial rolling speed and pressure parameter values ​​set for different types of tea leaves with different ages under standard temperature and humidity conditions, according to the three stages of rolling: before, during and after rolling.

[0118] The dynamic correction scheme is designed to address deviations in the state of tea leaves at different stages of the rolling process, and sets the trigger conditions and adjustment range for parameter adjustments.

[0119] For example, when the real-time humidity is more than 10% higher than the standard humidity threshold, a correction instruction to reduce the kneading pressure by 5%-8% is triggered for the middle stage of green tea kneading; when the initial tea strip regularity coefficient is less than 20% of the benchmark value, a correction instruction to increase the kneading speed by 3%-5% is triggered for the early stage of black tea kneading.

[0120] It should be noted that the purpose of this module is to eliminate the interference of environmental factors on the rolling process through multi-dimensional data comparison, screen out the universal and targeted rolling parameter adaptation rules, and generate personalized fine-tuning schemes. This can cover the rolling needs of tea leaves of different categories, different initial states, and different environmental conditions, thereby improving the stability and consistency of the quality of finished tea.

[0121] The technical solution of this invention is as follows:

[0122] This invention uses industrial vision sensors to collect real-time images of tea leaves during the rolling process, extracting core data such as strip formation rate, compactness, and sap overflow. These data are quantified into strip regularity coefficients, compactness coefficients, and sap overflow coefficients. A dynamic dataset is constructed by combining this data with current rolling speed and pressure parameters. Simultaneously, based on the practical experience of rolling masters, a tea rolling state database is established, categorized by tea type and rolling stage. The dynamic dataset is compared with the optimal parameters in the database to quantitatively assess the deviation between the current rolling state and the optimal working condition, clarifying the direction for parameter adjustment. This solves the problem of traditional rolling relying on manual experience and lacking objective data. The issues of state assessment and quantitative basis provide a precise time-related data foundation for subsequent adaptive control, ensuring the pertinence and scientific nature of parameter adjustments. Based on the direction of parameter adjustment, a two-layer control strategy of initial adjustment + dynamic fine-tuning is adopted. Combined with the tea state change trend prediction model, adaptive optimization of rolling parameters is achieved. At the same time, the parameter adjustment effect, real-time environmental temperature and humidity fluctuation information and historical optimal working condition data are integrated to screen the rolling parameter adaptation rules of different types of tea in specific temperature and humidity ranges. A personalized fine-tuning scheme including step-by-step parameter adjustment and temperature and humidity linkage compensation is generated, which improves the stability of tea rolling quality.

[0123] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A machine vision-based intelligent control system for tea leaf kneading, characterized in that: Includes the following modules: Tea leaf state extraction module: Extracts tea leaf state data during the rolling process, obtains current rolling parameters, and constructs a dynamic dataset; Adaptation Analysis and Adjustment Module: Obtain rolling parameters under different tea conditions, establish a tea rolling state database, input dynamic datasets into the tea rolling state database for adaptation analysis, quantify and evaluate the difference between the current rolling state and the optimal working condition through feature deviation, determine whether the rolling parameters need to be adjusted, and if so, confirm the adjustment direction; Adaptive adjustment module: Based on the adjustment direction of the kneading parameters, it performs adaptive kneading parameter adjustment, and predicts the trend of tea state change based on dynamic dataset, and dynamically adjusts the kneading parameters to adapt to the changes in tea state. Comparison and screening module: Obtain tea characteristic data after adjusting the rolling parameters, combine it with real-time environmental temperature and humidity fluctuation information, compare it with dynamic datasets and historical best working condition characteristics, screen the rolling parameter adaptation rules of different types of tea, and generate personalized fine-tuning schemes.

2. The intelligent control system for tea leaf kneading based on machine vision according to claim 1, characterized in that: The tea condition data includes the tea leaf strip ratio, tightness, and leaf juice leakage: Obtain the current tea leaf state image, extract data from the tea leaf state image, and obtain tea leaf state data; Image processing algorithms are used to extract tea state data from the obtained tea state images, and a dynamic dataset is constructed.

3. The intelligent control system for tea leaf kneading based on machine vision according to claim 2, characterized in that: Building dynamic datasets: Extract the tea leaf contour from the tea leaf condition image, obtain the tea leaf region for each tea leaf, calculate the aspect ratio and curvature based on the tea leaf region, and perform standardization processing. The aspect ratio and curvature are processed to obtain the strip formation coefficient of each tea leaf. The strip formation coefficients of all tea leaves in the tea leaf state image are averaged to obtain the strip regularity coefficient. Based on images of tea leaves, the contour compactness and surface texture uniformity of tea leaves are calculated. The compactness coefficient of each tea leaf is obtained by multiplying the contour compactness and the surface texture uniformity. The compactness coefficient of all tea leaves in the tea leaf state image is then averaged to obtain the compactness coefficient. Based on the tea leaf regions already acquired in the tea leaf condition image, the color saturation difference and surface highlight ratio of the tea leaves are calculated and then standardized. The color saturation difference and the surface highlight ratio are summed to obtain the juice overflow coefficient of each tea leaf. The juice overflow coefficients of all tea leaves in the tea leaf state image are averaged to obtain the leaf juice overflow degree coefficient. Obtain the current kneading parameters, including kneading speed and kneading pressure; By integrating the regularity coefficient of the strip, the compactness coefficient, the sap overflow coefficient, and the rolling parameters according to the same timestamp, a dynamic dataset is obtained.

4. The intelligent control system for tea kneading based on machine vision according to claim 1, characterized in that: Establish a database of tea rolling status: Obtain appropriate rolling parameters for different tea conditions, and archive the data in layers according to two dimensions: tea type and rolling stage; First, by classifying tea varieties, the suitable rolling parameters for tea of ​​the same variety are summarized. Then, the rolling stages are divided according to the suitable rolling parameters for tea of ​​the same variety. Based on the different rolling stages of tea during the rolling process, the suitable rolling parameters for different tea states are matched one by one, and a tea rolling state database is constructed.

5. The intelligent control system for tea kneading based on machine vision according to claim 1, characterized in that: Perform compatibility analysis: Input the tea state data from the dynamic dataset into the tea rolling state database, and output the corresponding adaptive rolling parameters. The adaptive kneading parameters and the kneading parameters are quantified separately to obtain the adaptive kneading parameter data and the kneading parameter data. The difference between the adapted kneading parameter data and the kneading parameter data is processed to obtain the parameter deviation value; If the parameter deviation value is greater than or equal to the upper limit of the deviation range, or less than or equal to the lower limit of the deviation range, then the current processing parameters are not suitable for the current tea rolling and need to be adjusted.

6. The intelligent control system for tea leaf kneading based on machine vision according to claim 1, characterized in that: Confirm the direction of adjustment: If the parameter deviation value is greater than or equal to the upper limit of the deviation range, the current kneading parameter exceeds the suitable kneading parameter under the same product category and kneading stage, and the current kneading parameter needs to be adjusted downward. If the parameter deviation value is less than or equal to the lower limit of the deviation range, the current kneading parameter is lower than the suitable kneading parameter for the same product category and kneading stage, and the current kneading parameter needs to be adjusted upward.

7. The intelligent control system for tea kneading based on machine vision according to claim 1, characterized in that: Adjust adaptive kneading parameters: Based on the adjustment direction and parameter deviation value of the kneading parameters, the initial adjustment of the kneading parameters is carried out, and the initial adjustment amount is the parameter deviation value; When the adjustment direction is downward, that is, when the parameter deviation value is greater than or equal to the upper limit of the deviation range, the corresponding kneading parameter is adjusted downward according to the initial adjustment. When the adjustment direction is upward, that is, when the parameter deviation value is less than or equal to the lower limit of the deviation range, the corresponding kneading parameter is adjusted upward according to the initial adjustment.

8. The intelligent control system for tea kneading based on machine vision according to claim 1, characterized in that: Predict the trend of tea leaf condition changes based on dynamic datasets: Extract tea state data with consecutive timestamps from dynamic datasets, calculate the slope of change of the regularity coefficient of strips, the compactness coefficient, and the degree of sap overflow coefficient within the same time period, and construct a tea state change trend prediction model through linear fitting. Based on the tea state change trend prediction model, after initial adjustment, the tea state data within the time period after initial adjustment is obtained. The tea state data after initial adjustment is input into the tea state change trend prediction model, and the predicted values ​​of the change direction and change magnitude of each state coefficient of tea in the next time period are output to predict the tea state change trend.

9. The intelligent control system for tea kneading based on machine vision according to claim 1, characterized in that: Dynamically adjust the kneading parameters to adapt to changes in the tea's state: Based on the predicted trend of tea state changes, the kneading parameters after the initial adjustment are dynamically adapted to the changes in tea state. If the predicted regularity coefficient of the tea leaves shows a continuous upward trend, and the current regularity coefficient of the tea leaves is close to the threshold of the optimal range for the same category and stage, then reduce the adjustment range of the rolling speed to avoid excessive strip formation of tea leaves leading to an increase in the broken leaf rate. If the predicted rate of increase in the compactness coefficient is slowing down, and the current compactness coefficient is below the optimal range threshold, then the rolling pressure adjustment range should be increased to enhance the compactness of the tea leaves.

10. The intelligent control system for tea kneading based on machine vision according to claim 1, characterized in that: Compare with dynamic datasets and historical best operating condition features: Acquire adaptively adjusted tea state data and real-time environmental temperature and humidity fluctuation information during the rolling process; The adaptively adjusted tea characteristic data and real-time environmental temperature and humidity fluctuation information are compared and analyzed in multiple dimensions with the historical best working condition characteristic data of the same category and the same kneading stage in the tea kneading state database. Construct a multi-dimensional comparison and evaluation index system, which includes feature similarity, working condition matching degree, and environmental adaptability; The feature similarity, working condition matching degree, and environmental adaptability are summed to obtain the comprehensive matching coefficient; If the overall matching coefficient is greater than or equal to the overall matching threshold, it means that the adjusted kneading parameters and the corresponding environment are suitable for the current kneading stage of the corresponding tea category. The kneading parameters, environmental temperature and humidity conditions, and tea characteristic data are archived according to the category-stage-environment dimension to obtain the kneading parameter adaptation rules.

Images