A rolling parameter dynamic adjustment method, system, medium and device

Abstract

The application discloses a kind of kneading parameter dynamic adjustment method, system, medium and equipment, belong to tea processing technical field, method includes: acquisition original tea leaf characteristic information, after original tea leaf characteristic information pre-processing, through three-dimensional point cloud reconstruction extraction leaf vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics, combined with hyperspectral analysis obtains cell layer light transmission characteristics;Correlation kneading stage information extracts dynamic mechanical characteristics, establishes the mechanical characteristic code library containing typical kneading mode;Mapping relationship network of tea variety and optimal pressure gradient distribution, kneading rhythm mode is constructed, and initial adjustment strategy is generated based on the physical property conversion relationship function of benchmark tea and target tea green;Collect real-time state information, dynamically optimize initial adjustment strategy.The application realizes the self-adaptive precision regulation and control of kneading process by the dynamic correlation of quantitative tea physical properties and mechanical response, significantly improves the standardization level and finished product quality of tea processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tea processing, and in particular to a rolling parameter dynamic adjustment method, system, medium and equipment. BACKGROUND

[0002] Tea rolling is an important process for shaping tea leaves and promoting the transformation of internal substances. The accurate control of its process parameters directly affects the tightness of tea strips, cell breakage rate and flavor substance release degree. The traditional rolling process mainly relies on manual experience to judge the pressure and rolling time, which can meet the production needs to some extent. However, due to the lack of systematic detection and analysis of tea physical properties, the setting of process parameters often cannot accurately match the actual characteristics of different varieties and different batches of raw materials. Especially when facing fluctuations in raw material quality, fixed processing parameters cannot respond to the dynamic changes of tea properties in the processing process in real time, which easily leads to unstable product quality and restricts the improvement space of product quality to some extent. In addition, due to the lack of effective online monitoring means, the process adjustment has a lag, which easily leads to problems such as insufficient cell breakage or excessive pressure damage to tea polyphenol structure. These problems make the traditional rolling process face challenges in meeting the production needs of high-quality tea. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a rolling parameter dynamic adjustment method, system, medium and equipment, which realizes the adaptive control of the rolling process by establishing the dynamic mapping relationship between tea physical properties and mechanical response and real-time monitoring of the cell breakage state, and solves the problem of unstable quality caused by fixed parameters and experience dependence in the traditional rolling process.

[0004] In order to achieve the above technical purpose, in a first aspect, the present application provides a rolling parameter dynamic adjustment method, comprising:

[0005] Collecting original tea property information, the original tea property information including first leaf image information, first response information and first withering information;

[0006] Pretreating the original tea property information to obtain processed tea property information, the processed tea property information including second leaf image information, second response information and second withering information after process alignment;

[0007] Reconstructing the second leaf image information in three-dimensional point cloud, extracting static physical parameters, the static physical parameters including vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics, and performing hyperspectral analysis on the second leaf image information to obtain cell layer light transmission characteristics;

[0008] The second response information and the second wilting information are associated with the rubbing stage information, dynamic mechanical characteristics are extracted, the dynamic mechanical characteristics include elastic deformation recovery rate and cell wall rupture threshold, a mechanical characteristic code library containing typical rubbing modes is established, and the typical rubbing modes include light pressure shaping, medium pressure promoting rupture and heavy pressure shaping.

[0009] According to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics and the mechanical characteristic code library, a mapping relationship network of different tea varieties and the best pressure gradient distribution and rubbing rhythm mode is constructed.

[0010] Based on the static physical parameters and the dynamic mechanical characteristics, a physical property conversion relationship function of the reference tea and the target tea is established.

[0011] According to the mapping relationship network and the physical property conversion relationship function, an initial adjustment strategy is generated and executed, and the initial adjustment strategy includes pressure gradient distribution parameters, rubbing rhythm parameters and rubbing time length parameters.

[0012] Real-time state information is collected in the rubbing process, and the real-time state information includes cell damage rate information and red edge formation degree information.

[0013] The real-time state information is compared with the predicted state information in the initial adjustment strategy, and the pressure gradient distribution parameters, the rubbing rhythm parameters and the rubbing time length parameters are dynamically adjusted according to the comparison result, an optimized adjustment strategy is generated and updated.

[0014] In some embodiments, the second leaf image information is reconstructed into a three-dimensional point cloud, and static physical parameters are extracted, the static physical parameters include vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics, including:

[0015] Multi-view point cloud information of the leaf surface is obtained, and the multi-view point cloud information includes overhead angle point cloud, side angle point cloud and oblique angle point cloud.

[0016] The multi-view point cloud information is registered and fused, the registration and fusion processing includes using an iterative closest point algorithm to unify the overhead angle point cloud, the side angle point cloud and the oblique angle point cloud to the same coordinate system, and realizing point cloud density equalization through voxel grid filtering to obtain registered point cloud data.

[0017] Based on the registered point cloud data, a three-dimensional geometric model of the leaf surface is reconstructed, and the three-dimensional geometric model of the leaf surface is configured to generate a closed surface by using a Poisson surface reconstruction algorithm and to apply Laplace smoothing to eliminate surface noise.

[0018] In the three-dimensional geometric model of the leaf surface, the point cloud Gaussian curvature distribution is calculated, the curvature maximum points are extracted to constitute the main ridge line contour of the leaf vein, and the continuous ridge line trajectory is obtained by B-spline curve fitting to obtain the main ridge line of the leaf vein.

[0019] A local coordinate system is established in the normal plane of the ridge trajectory of the main ridge line of the vein, the vertical distance of each sampling point from the reference surface is measured, a height-position distribution function is generated, and a convex height distribution curve is extracted, denoted as vein convex height information;

[0020] In the three-dimensional geometric model of the blade surface, equidistant measurement points are set along the blade edge contour line, the blade thickness at each point is calculated by the ray method, and the thickness gradient between adjacent points is calculated to obtain a thickness gradient change rate, denoted as blade edge thickness information;

[0021] A spatial correlation matrix of the vein convex height information and the blade edge thickness information is established, including:

[0022] The vein height distribution curve and the blade edge thickness gradient sequence are spatio-temporally aligned;

[0023] A two-dimensional feature mapping table reflecting the correlation between the two is constructed to obtain a spatial correlation matrix;

[0024] Static physical parameters are extracted in the spatial correlation matrix.

[0025] In some embodiments, the second leaf image information is subjected to hyperspectral analysis to obtain cell layer light transmission characteristics, including:

[0026] A characteristic spectral curve of the mesophyll region is extracted from the second leaf image information;

[0027] In the characteristic spectral curve, a first spectral reflectance at a 1450 nm water molecule absorption characteristic band and a second spectral reflectance at a 1940 nm water molecule absorption characteristic band are selected;

[0028] The absorption depth ratio of the first spectral reflectance and the second spectral reflectance is calculated to obtain a light transmission characteristic conversion coefficient;

[0029] The characteristic spectral curve is spatially registered with the three-dimensional geometric model of the blade surface;

[0030] A light transmission rate calculation model is established according to the light transmission characteristic conversion coefficient, and the registered three-dimensional geometric model is converted into a cell layer light transmission characteristic spatial distribution map;

[0031] The cell layer light transmission characteristic is extracted from the cell layer light transmission characteristic spatial distribution map.

[0032] In some embodiments, the second response information and the second wilting information are associated with rolling stage information to extract dynamic mechanical characteristics, including:

[0033] Pressure-deformation characteristic parameters are obtained from the second response information, including an elastic deformation stage slope and a plastic yield point position;

[0034] extracting a wilting degree index from the second wilting information, the wilting degree index including a leaf water content change gradient;

[0035] coupling analysis of the elastic deformation stage slope and the wilting degree index to generate an elastic deformation recovery rate, including:

[0036] establishing a decay function of the elastic deformation stage slope with the wilting degree;

[0037] calculating the slope offset at the current wilting degree in the decay function;

[0038] normalizing the slope offset to generate the elastic deformation recovery rate;

[0039] associating the plastic yield point position with the wilting degree index to obtain a cell wall rupture threshold, including:

[0040] constructing a nonlinear regression model of the yield point displacement and the wilting degree;

[0041] interpolating calculation of the expected yield point position in the nonlinear regression model according to the current wilting degree;

[0042] calculating the position deviation value from the measured yield point position and the expected yield point position;

[0043] converting the position deviation value into the cell wall rupture threshold;

[0044] and, establishing a mechanical characteristic coding library containing typical rolling modes, including:

[0045] dividing the elastic deformation recovery rate into three characteristic intervals of light pressure shaping, medium pressure promoting rupture, and heavy pressure shaping;

[0046] setting the pressure action time length of each characteristic interval based on the cell wall rupture threshold;

[0047] constructing a correspondence table of the characteristic interval and the rolling process parameters according to the pressure action time length to obtain the mechanical characteristic coding library.

[0048] In some embodiments, according to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics, and the mechanical characteristic coding library, a mapping relationship network of different tea varieties and the best pressure gradient distribution and rolling rhythm mode is constructed, including:

[0049] spatially registering the vein protrusion height distribution characteristics and the elastic deformation recovery rate to generate a regional pressure sensitivity distribution map;

[0050] correlation analysis of the cell layer light transmission characteristics and the cell wall rupture threshold to establish a light transmission characteristic-pressure resistance strength correspondence table;

[0051] Match the typical rolling mode parameter combination of tea varieties based on the feature interval in the mechanical feature coding library;

[0052] According to the light transmission characteristic-pressure resistance corresponding relationship table and the typical rolling mode parameter combination, a mapping relationship network is constructed;

[0053] Based on the static physical parameters and dynamic mechanical characteristics, a physical property conversion relationship function between the reference tea and the target tea is established, including:

[0054] The reference leaf margin thickness gradient feature of the reference tea and the target leaf margin thickness gradient feature of the target tea are extracted, and a preset thickness conversion coefficient is calculated;

[0055] According to the preset thickness conversion coefficient, a preset elastic deformation recovery rate is obtained, and an elastic response correction factor is generated according to the preset elastic deformation recovery rate and the real-time elastic deformation recovery rate;

[0056] According to the preset thickness conversion coefficient and the elastic response correction factor, the physical property conversion relationship function is established.

[0057] In some embodiments, according to the mapping relationship network and the physical property conversion relationship function, an initial adjustment strategy is generated and executed, the initial adjustment strategy including pressure gradient distribution parameters, rolling rhythm parameters and rolling time length parameters, including:

[0058] According to the regional pressure sensitivity distribution map, an initial pressure gradient distribution parameter is generated, the initial pressure gradient distribution parameter including a first pressure value of the vein area and a second pressure value of the mesophyll area;

[0059] The pressure resistance threshold of the current tea variety is queried from the light transmission characteristic-pressure resistance corresponding relationship table;

[0060] According to the pressure resistance threshold, the rolling rhythm parameters are set, the rolling rhythm parameters including a first duration of the pressure increasing stage and a second duration of the pressure releasing stage;

[0061] Based on the typical rolling mode parameter combination, a basic rolling time length parameter is determined;

[0062] The basic rolling time length parameter is corrected through the physical property conversion relationship function to generate an actual rolling time length parameter adapted to the current tea;

[0063] The initial pressure gradient distribution parameter, the rolling rhythm parameter and the actual rolling time length parameter are combined to form the initial adjustment strategy;

[0064] According to the initial adjustment strategy, the rolling equipment is controlled to execute the initial rolling process.

[0065] In some embodiments, the real-time state information is compared with the predicted state information in the initial adjustment strategy, and the pressure gradient distribution parameters, the rolling rhythm parameters and the rolling duration parameters are dynamically adjusted according to the comparison results, an optimized adjustment strategy is generated and updated, including:

[0066] The current cell damage rate and the current red edge formation degree in the real-time state information are obtained;

[0067] The current cell damage rate is compared with the expected cell damage rate in the initial adjustment strategy, and a damage rate deviation value is calculated;

[0068] The current red edge formation degree is compared with the expected red edge formation degree in the initial adjustment strategy, and a red edge deviation value is calculated;

[0069] The pressure gradient distribution parameters are adjusted according to the damage rate deviation value, including:

[0070] When the damage rate deviation value is negative, the first pressure value and the second pressure value are increased by a preset gradient;

[0071] When the damage rate deviation value is positive, the first pressure value and the second pressure value are decreased by a preset proportion;

[0072] The rolling rhythm parameters are adjusted according to the red edge deviation value, including:

[0073] When the red edge deviation value is negative, the first duration is extended and the second duration is shortened;

[0074] When the red edge deviation value is positive, the first duration is shortened and the second duration is extended;

[0075] The actual rolling duration parameters are recalculated based on the adjusted pressure gradient distribution parameters and the adjusted rolling rhythm parameters;

[0076] The adjusted pressure gradient distribution parameters, the adjusted rolling rhythm parameters and the recalculated actual rolling duration parameters are combined to form an optimized adjustment strategy;

[0077] The control parameters of the rolling device are updated according to the optimized adjustment strategy.

[0078] In a second aspect, the present application further provides a rolling parameter dynamic adjustment system suitable for the method of the first aspect, the system comprising a collection module, a data processing module and a logic operation module;

[0079] The collection module is used for collecting original tea leaf characteristic information, and the original tea leaf characteristic information includes first leaf image information, first response information and first withering information;

[0080] The data processing module is configured to preprocess original tea leaf characteristic information to obtain processed tea leaf characteristic information, and the processed tea leaf characteristic information includes the second leaf image information, the second response information and the second withering information after process alignment.

[0081] The logic operation module is configured to perform three-dimensional point cloud reconstruction on the second leaf image information, extract static physical parameters, and perform hyperspectral analysis on the second leaf image information to obtain cell layer light transmission characteristics, wherein the static physical parameters include vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics; perform information correlation of the second response information and the second withering information in the rolling and twisting stage, extract dynamic mechanical characteristics, and establish a mechanical characteristic code library containing typical rolling and twisting modes, wherein the dynamic mechanical characteristics include elastic deformation recovery rate and cell wall rupture threshold, and the typical rolling and twisting modes include light pressure shaping, medium pressure promoting rupture, and heavy pressure shaping; construct a mapping relationship network of different tea varieties and optimal pressure gradient distribution and rolling and twisting rhythm mode according to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics and the mechanical characteristic code library; establish a physical property conversion relationship function of reference tea and target tea according to the static physical parameters and the dynamic mechanical characteristics; generate and execute an initial adjustment strategy according to the mapping relationship network and the physical property conversion relationship function, wherein the initial adjustment strategy includes pressure gradient distribution parameters, rolling and twisting rhythm parameters and rolling and twisting time length parameters; collect real-time state information in the rolling and twisting process, wherein the real-time state information includes cell damage rate information and red edge formation degree information; compare the real-time state information with predicted state information in the initial adjustment strategy, dynamically adjust the pressure gradient distribution parameters, the rolling and twisting rhythm parameters and the rolling and twisting time length parameters according to the comparison result, generate an optimized adjustment strategy and update.

[0082] In a third aspect, the present application also provides a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method of the first aspect.

[0083] In a fourth aspect, the present application also provides an electronic device comprising a memory and a processor, wherein the memory is configured to store one or more computer program instructions, and the one or more computer program instructions are executed by the processor to implement the method of the first aspect.

[0084] Compared with the prior art, the application has the beneficial effects that the application provides a rolling parameter dynamic adjustment method, system, medium and equipment, the method comprising: collecting original tea leaf characteristic information, including first leaf image information, first response information and first withering information; after preprocessing the original tea leaf characteristic information, extracting vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics through three-dimensional point cloud reconstruction, and combining with hyperspectral analysis to obtain cell layer light transmission characteristics; correlating rolling stage information to extract dynamic mechanical characteristics, establishing a mechanical characteristic code library containing light pressure shaping, medium pressure breaking and heavy pressure shaping; constructing a mapping relationship network of tea varieties and optimal pressure gradient distribution and rolling rhythm mode, and generating an initial adjustment strategy based on the physical property conversion relationship function of the reference tea and the target tea; collecting cell damage rate and red edge formation degree information in real time, and dynamically optimizing pressure gradient distribution parameters, rolling rhythm parameters and rolling time length parameters. The application realizes self-adaptive and accurate control of the rolling process by dynamically correlating the quantitative tea physical properties and mechanical responses, solves the quality instability problem caused by the dependence of the traditional method on experience and the parameter solidification, and significantly improves the standardization level of tea processing and the quality of finished products. BRIEF DESCRIPTION OF DRAWINGS

[0085] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0086] Figure 1 is a method step diagram of steps S101 to S109 of the adjustment method described in the specific embodiment;

[0087] Figure 2 is a method step diagram of steps S201 to S207 of the adjustment method described in the specific embodiment;

[0088] Figure 3 is a structure schematic diagram of the electronic device described in the specific embodiment.

[0089] The reference signs are as follows:

[0090] 1. Electronic device;

[0091] 11. Memory;

[0092] 12. Processor. DETAILED DESCRIPTION

[0093] The application will be described in further detail below with reference to the drawings and embodiments. It is particularly pointed out that the following embodiments are only for illustration of the application, but do not limit the scope of the application. Similarly, the following embodiments are only part of the embodiments of the application, and all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0094] Please refer to Figure 1 In the first aspect, the embodiment provides a rolling parameter dynamic adjustment method, comprising:

[0095] S101, collecting original tea leaf characteristic information, the original tea leaf characteristic information comprising first leaf image information, first response information and first withering information;

[0096] S102, preprocessing the original tea leaf characteristic information to obtain processed tea leaf characteristic information, the processed tea leaf characteristic information comprising second leaf image information, second response information and second withering information after process alignment;

[0097] S103, performing three-dimensional point cloud reconstruction on the second leaf image information to extract static physical parameters, the static physical parameters comprising vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics, and performing hyperspectral analysis on the second leaf image information to obtain cell layer light transmission characteristics;

[0098] S104, correlating the second response information and the second withering information to extract dynamic mechanical characteristics, the dynamic mechanical characteristics comprising elastic deformation recovery rate and cell wall rupture threshold, and establishing a mechanical characteristic code library containing typical rolling modes, the typical rolling modes comprising light pressure shaping, medium pressure promoting rupture and heavy pressure shaping;

[0099] S105, constructing a mapping relationship network of different tea varieties and optimal pressure gradient distribution and rolling rhythm mode according to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics and the mechanical characteristic code library;

[0100] S106, establishing a physical property conversion relationship function of reference tea and target tea based on the static physical parameters and the dynamic mechanical characteristics;

[0101] S107, generating and executing an initial adjustment strategy according to the mapping relationship network and the physical property conversion relationship function, the initial adjustment strategy comprising pressure gradient distribution parameters, rolling rhythm parameters and rolling time length parameters;

[0102] S108, collecting real-time state information in the rolling process, the real-time state information comprising cell damage rate information and red edge formation degree information;

[0103] S109, compare the real-time state information with the predicted state information in the initial adjustment strategy, dynamically adjust the pressure gradient distribution parameter, the rolling rhythm parameter and the rolling time length parameter according to the comparison result, generate an optimized adjustment strategy and update.

[0104] In step S101, the collection of original tea leaf characteristic information can adopt a multi-modal sensing fusion technology. Preferably, the first leaf image information is acquired by a dual-spectrum acquisition system combining visible light and near-infrared imaging, which can capture surface morphological features and preliminarily judge internal moisture distribution; the first response information is measured by an embedded micro-force sensor array, and the stress-strain curve of the tea leaf is acquired under the pre-pressing condition simulating the actual rolling process; and the first withering information is integrated with a temperature and humidity sensor and a near-field communication technology to realize the dynamic correlation between the withering environment parameters and the tea leaf properties. This synchronous collection mechanism of multi-source heterogeneous data provides a more comprehensive feature dimension for subsequent analysis.

[0105] In step S102, preferably, the preprocessing process introduces a time alignment algorithm to solve the clock drift problem between different sampling devices through dynamic time warping, ensuring that the texture changes in the second leaf image information and the mechanical curve of the second response information are accurately matched in the time domain; in particular, the second withering information is processed by a sliding window standardization to eliminate the systematic deviation of withering degree evaluation between batches and improve the reliability of subsequent feature extraction.

[0106] In step S103, preferably, the three-dimensional point cloud reconstruction adopts a method combining structured light projection and phase measurement profilometry, and the extraction of the leaf vein protrusion height distribution feature introduces morphological gradient operation to accurately quantify the three-dimensional morphological differences between the main veins and the side veins; the leaf edge thickness gradient feature is represented by calculating the curvature change rate of the point cloud data. Preferably, the hyperspectral analysis adopts a feature band selection algorithm, focusing on specific absorption peaks related to cell wall structure, making the detection of cell layer light transmission characteristics more targeted. This step realizes comprehensive characterization from macroscopic morphology to microscopic structure through a multi-scale feature fusion method.

[0107] In step S104, preferably, the extraction of dynamic mechanical characteristics adopts a time-frequency joint analysis method, and the elastic deformation recovery rate is calculated by decomposing the intrinsic mode function from the response signal through Hilbert-Huang transform; the cell wall rupture threshold combines the acoustic emission detection technology, and when a characteristic frequency acoustic signal mutation occurs, it is determined as a critical rupture point. The construction of the mechanical characteristic code library introduces a transfer learning strategy to map the feature space of typical rolling modes to a unified measurement standard, enhancing the process transferability between different varieties of tea leaves.

[0108] In step S105, the construction of the mapping relationship network can adopt a graph neural network architecture, taking static physical parameters as node features and dynamic mechanical features as edge weights, and learning the nonlinear relationship between different features through a graph attention mechanism. Further, an adversarial sample enhancement technique is introduced during network training to improve the model's generalization ability by simulating raw material fluctuations, ensuring stable parameter recommendation performance in the face of actual production quality fluctuations.

[0109] In step S106, the physical property conversion relationship function preferably establishes a differential homeomorphism mapping between the reference tea and the target tea feature spaces through a local linear embedding method. This function takes into account the geodesic distance of different varieties of tea on the feature manifold, allowing the parameter conversion process to maintain the topological structure of tea properties and improve the accuracy of cross-variety process adjustment.

[0110] In step S107, the generation of the initial adjustment strategy introduces a multi-objective optimization framework, with the pressure gradient distribution parameter seeking a balance between cell damage rate and filament integrity through Pareto frontier analysis; the rolling rhythm parameter optimizes the timing combination of pressure and pressure release using a dynamic programming method; and the rolling duration parameter is adaptively determined based on the convergence speed of the target point in the feature space, ensuring the coordinated optimization of the initial adjustment strategy in multiple quality indicators.

[0111] In step S108, preferably, real-time state monitoring combines microscopic vision and spectral imaging technology, with cell damage rate information enhanced by fluorescence staining for improved detection sensitivity, and red edge formation degree information quantified using feature spectra after principal component analysis dimensionality reduction. The monitoring system uses an edge computing architecture to complete feature extraction at the data acquisition end, reducing information transmission delay.

[0112] In step S109, the dynamic adjustment algorithm is a hierarchical control structure, with the upper layer generating parameter adjustment directions based on a model predictive control framework, and the lower layer using an adaptive PID controller to achieve precise adjustment of the actuator. In particular, the optimization strategy update introduces an incremental learning method, adjusting only the feature dimensions with significant deviations, ensuring response speed while avoiding process fluctuations caused by excessive adjustment. Through hierarchical and progressive regulation, the entire system has good anti-interference ability and stability.

[0113] The working principle of the embodiment can be understood as follows: original tea leaf characteristic information is collected through a multi-modal sensor fusion technology, preprocessed through time sequence alignment, combined with three-dimensional point cloud reconstruction and hyperspectral analysis to extract static physical parameters and cell layer light transmission characteristics, and a mechanical characteristic code library containing typical rolling modes is established based on information correlation in the rolling stage to construct a mapping relationship network of tea varieties and optimal processing parameters. The embodiment establishes an accurate physical property conversion function through the cooperative analysis of static physical parameters and dynamic mechanical characteristics, significantly improves the process adaptability of different batches of raw materials; based on the dynamic comparison mechanism of real-time state information and predicted state, the pressure gradient distribution parameters, rolling rhythm parameters and rolling time parameters can be adjusted adaptively, effectively ensuring the stability of the processing quality; through the joint application of the mechanical characteristic code library and the mapping relationship network, the intelligent matching and optimization of the typical rolling mode are realized, providing a standardized and intelligent solution for the tea rolling process.

[0114] Please refer to Figure 2 In some embodiments, the second leaf image information is subjected to three-dimensional point cloud reconstruction to extract static physical parameters, including vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics, including:

[0115] S201, acquiring multi-view point cloud information of a leaf surface, the multi-view point cloud information including overhead angle point cloud, side angle point cloud and oblique angle point cloud;

[0116] S202, performing registration and fusion processing on the multi-view point cloud information, the registration and fusion processing including using an iterative closest point algorithm to unify the overhead angle point cloud, the side angle point cloud and the oblique angle point cloud to the same coordinate system, and realizing point cloud density equalization through voxel grid filtering to obtain registered point cloud data;

[0117] S203, reconstructing a three-dimensional geometric model of the leaf surface based on the registered point cloud data, the three-dimensional geometric model of the leaf surface being configured to generate a closed surface using a Poisson surface reconstruction algorithm and to apply Laplace smoothing to eliminate surface noise;

[0118] S204, in the three-dimensional geometric model of the leaf surface, calculating a point cloud Gaussian curvature distribution, extracting curvature maximum points to constitute a vein main ridge line contour, and obtaining a continuous ridge line trajectory through B-spline curve fitting to obtain a vein main ridge line;

[0119] S205, establishing a local coordinate system in the normal plane of the ridge line trajectory of the vein main ridge line, measuring the vertical distance of each sampling point from the reference surface, generating a height-position distribution function, and extracting a protrusion height distribution curve, denoted as vein protrusion height information;

[0120] S206, setting equidistant measurement points along the blade edge contour line in the three-dimensional geometric model of the blade surface, calculating the blade thickness at each point by the ray method, and calculating the thickness gradient between adjacent points to obtain the thickness gradient change rate, denoted as the blade edge thickness information;

[0121] S207, establishing a spatial correlation matrix of the vein ridge height information and the blade edge thickness information, including:

[0122] spatiotemporal alignment of the vein height distribution curve and the blade edge thickness gradient sequence;

[0123] constructing a two-dimensional feature mapping table reflecting the correlation between the two to obtain the spatial correlation matrix;

[0124] extracting static physical parameters in the spatial correlation matrix.

[0125] In step S201, the multi-view point cloud information is a set of leaf surface spatial coordinate data collected from different directions by a three-dimensional scanning device. The overhead view point cloud represents the morphological characteristics of the upper surface of the leaf, the side view point cloud is used to capture the contour information of the side edge of the leaf, and the oblique view point cloud is obtained by 45-degree inclined scanning to obtain the geometric characteristics of the transition region of the leaf. The point cloud data of the three views together constitute the original observation data of the complete three-dimensional morphology of the leaf.

[0126] In step S202, the registration and fusion process realizes the spatial alignment of multi-source point clouds by the iterative closest point algorithm, which unifies the overhead view point cloud, the side view point cloud and the oblique view point cloud to the same coordinate system by minimizing the distance between corresponding points; voxel grid filtering is used for point cloud density equalization, which establishes a three-dimensional voxel grid and retains the center of gravity point in each voxel to eliminate the uneven distribution of point clouds caused by the difference in scanning angle. The registered point cloud data has spatial consistency and uniform density distribution, laying a foundation for subsequent three-dimensional reconstruction.

[0127] In step S203, the three-dimensional geometric model of the leaf surface is generated by the Poisson surface reconstruction algorithm, which uses the point cloud normal vector information to construct an indicator function and extract an isosurface to form a closed leaf surface model; Laplace smoothing processing eliminates surface noise generated in the reconstruction process by iterative optimization of vertex position, so that the model maintains both geometric details and smooth transition characteristics. The reconstructed geometric model accurately reflects the micro surface morphology of the leaf.

[0128] In step S204, the vein ridge line extraction process first calculates the Gaussian curvature distribution of the point cloud, and the curvature maximum point corresponds to the convex region of the vein; B-spline curve fitting constructs a smooth and continuous ridge line trajectory by optimizing the control points, which accurately reflects the three-dimensional spatial trend of the main vein. The accuracy of ridge line extraction directly affects the accuracy of subsequent height measurement.

[0129] In step S205, the establishment of the local coordinate system is based on the ridge line trajectory, and the height measurement in the normal plane is achieved by calculating the vertical distance between the sampling points and the reference plane; the height-position distribution function completely records the protrusion degree of each part of the main vein, and the protrusion height distribution curve is the quantitative information of the vein protrusion height. This parameter reflects the physical property differences of tea raw materials.

[0130] In step S206, the leaf margin thickness is measured by the ray method, that is, normal rays are emitted at equally spaced measurement points on the leaf margin contour line, and the penetration distance is calculated; the thickness gradient is calculated by the thickness difference between adjacent points, and the thickness gradient change rate is the thickness information of the leaf margin. This parameter is closely related to the mechanical response in the rolling process.

[0131] In step S207, the construction of the spatial correlation matrix establishes a mapping relationship between the vein height distribution curve and the leaf margin thickness gradient sequence through space-time alignment; the two-dimensional feature mapping table stores the correlation data of the two in the form of a matrix, and the rows and columns of the matrix correspond to the height and thickness feature dimensions respectively. The static physical parameters are extracted from the matrix, which provides a feature basis for subsequent process parameter optimization.

[0132] The present embodiment realizes the accurate quantification of the static physical parameters of tea leaves through multi-view point cloud fusion and three-dimensional reconstruction technology. Among them, the vein protrusion height distribution feature reflects the mechanical bearing capacity of the raw materials, and the leaf margin thickness gradient feature is related to the deformation characteristics in the rolling process. The establishment of the spatial correlation matrix forms a unified evaluation system for these two key parameters, providing a reliable feature basis for process adaptation. The present embodiment ensures the accuracy of parameter extraction through a strict point cloud processing process, and improves the quality control basis of tea processing from the three-dimensional geometric feature level.

[0133] In some embodiments, the second leaf image information is subjected to hyperspectral analysis to obtain the light transmission characteristics of the cell layer, including:

[0134] Extracting a characteristic spectral curve of the mesophyll region in the second leaf image information;

[0135] Selecting a first spectral reflectance of a 1450 nm water molecule absorption characteristic band and a second spectral reflectance of a 1940 nm water molecule absorption characteristic band in the characteristic spectral curve;

[0136] Calculating the absorption depth ratio of the first spectral reflectance and the second spectral reflectance to obtain a light transmission characteristic conversion coefficient;

[0137] Spatially registering the characteristic spectral curve with a three-dimensional geometric model of the leaf surface;

[0138] Establishing a light transmission rate calculation model according to the light transmission characteristic conversion coefficient, and converting the registered three-dimensional geometric model into a spatial distribution map of the cell layer light transmission characteristics;

[0139] extracting the light transmission characteristics of the cell layer from the spatial distribution map of the light transmission characteristics of the cell layer.

[0140] In the present embodiment, the second leaf image information is hyperspectral image data that has been pre-processed, such as radiation correction and geometric correction, and the characteristic spectral curve of the mesophyll region is a typical spectral response extracted from the second leaf image information. After excluding the vein region by manual annotation or automatic segmentation algorithm, the average reflectance curve of the selected mesophyll region pixels is calculated to obtain.

[0141] The 1450 nm water molecule absorption characteristic band is the first characteristic band, which corresponds to the O-H bond first-order frequency absorption peak of water molecules, and the first spectral reflectance value reflects the absorption degree of near-infrared light by water molecules inside the leaf. The 1940 nm water molecule absorption characteristic band is the second characteristic band, which corresponds to the O-H bond combination frequency absorption peak of water molecules, and the second spectral reflectance value represents the more intense absorption characteristics of water molecules. The selection of the first characteristic band and the second characteristic band is based on the typical absorption characteristics of water molecules in the near-infrared region.

[0142] The light transmission characteristic conversion coefficient is a quantitative parameter obtained by calculating the absorption depth ratio of the first spectral reflectance and the second spectral reflectance. The specific implementation is to take the ratio of the reflectance valley value of the characteristic band to the reflectance of the adjacent peak, and then eliminate the influence of light conditions by standardization processing. The light transmission characteristic conversion coefficient establishes a quantitative conversion relationship between the spectral absorption characteristics and the light transmission ability of the cell layer.

[0143] Spatial registration is the process of establishing a spatial correspondence between the characteristic spectral curve and the three-dimensional geometric model of the leaf surface. By matching the coordinate systems of the hyperspectral image and the three-dimensional model, it is ensured that each three-dimensional model surface point can be associated with the corresponding spectral feature data. During registration, the consistency of the spectral data and the geometric model in scale, rotation and translation parameters is maintained. Preferably, the registration process uses a combination of feature point matching and affine transformation.

[0144] The light transmission rate calculation model is a mathematical relationship established based on the light transmission characteristic conversion coefficient, which converts the spectral absorption characteristics into the light transmission rate value of the cell layer. The model construction considers the influence of leaf thickness, cell arrangement density and other factors on light transmission, and maps the absorption depth ratio to the actual light transmission rate value through an empirical formula. Further, the model parameters are obtained by standard sample calibration to ensure the physical accuracy of the conversion result.

[0145] The spatial distribution map of the light transmission characteristics of the cell layer is a pseudo-color image generated by applying a light transmission rate calculation model to the registered three-dimensional geometric model. Different color regions in the image directly show the differences in light transmission characteristics of different parts of the leaf. Preferably, a linear interpolation algorithm is used to ensure spatial continuity. The spatial resolution of the spatial distribution map of the light transmission characteristics of the cell layer is consistent with the original three-dimensional model, and can accurately reflect the micro changes in the light transmission characteristics on the surface of the leaf. The light transmission characteristic parameters of the cell layer extracted from the spatial distribution map of the light transmission characteristics of the cell layer include the area average light transmission rate and the light transmission uniformity index. These characteristic parameters reflect the structural integrity and arrangement state of the cell layer of the leaf, and provide a micro-level basis for evaluating the quality of tea raw materials.

[0146] The present embodiment ensures the reliability of the analysis basis through strict data preprocessing, uses the characteristic absorption band of water molecules to reflect the structural characteristics of the cell layer, and establishes a complete conversion chain from macro-spectrum to micro light transmission characteristics. The spatial registration link realizes the accurate fusion of two-dimensional spectrum and three-dimensional geometry, and the finally generated spatial distribution map of the light transmission characteristics of the cell layer can directly display the quality of the leaf cell layer, providing a key basis for subsequent process optimization. Through accurate analysis of the characteristic wave band of water molecules in the second leaf image information, the present embodiment establishes the spatial registration relationship between the light transmission characteristic conversion coefficient and the three-dimensional geometric model, realizes the nondestructive testing and visual expression of the light transmission characteristics of the cell layer, and provides an objective and quantitative microstructure basis for tea quality evaluation.

[0147] In some embodiments, the second response information and the second wilting information are associated with the rolling stage information, and dynamic mechanical characteristics are extracted, including:

[0148] Pressure-deformation characteristic parameters are obtained from the second response information, including the elastic deformation stage slope and the plastic yield point position;

[0149] A wilting degree index is extracted from the second wilting information, including a leaf water content change gradient;

[0150] The elastic deformation stage slope and the wilting degree index are coupled and analyzed to generate an elastic deformation recovery rate, including:

[0151] An attenuation function of the elastic deformation stage slope with respect to the wilting degree is established;

[0152] The slope offset at the current wilting degree is calculated in the attenuation function;

[0153] The slope offset is normalized to generate the elastic deformation recovery rate;

[0154] The plastic yield point position is associated with the wilting degree index to obtain a cell wall rupture threshold, including:

[0155] a nonlinear regression model of the yield point displacement and the wilting degree is constructed;

[0156] an expected yield point position is obtained by interpolation calculation in the nonlinear regression model according to the current wilting degree;

[0157] a position deviation value is obtained by deviation value calculation of the measured yield point position and the expected yield point position;

[0158] the position deviation value is converted into a cell wall rupture threshold value;

[0159] and a mechanical characteristic code library containing typical rolling and twisting modes is established, including:

[0160] three characteristic intervals of light pressure shaping, medium pressure promoting rupture and heavy pressure shaping are divided according to the elastic deformation recovery rate;

[0161] the pressure action time length of each characteristic interval is set based on the cell wall rupture threshold value;

[0162] a corresponding relationship table of the characteristic intervals and rolling and twisting process parameters is constructed according to the pressure action time length, and the mechanical characteristic code library is obtained.

[0163] In the embodiment, the second response information is dynamic data of force response of the leaf during rolling and twisting collected by a mechanical sensor, and the pressure-deformation characteristic parameter is used to represent the deformation characteristic of the leaf under mechanical force, wherein the slope in the elastic deformation stage reflects the ability of the leaf tissue to resist deformation, and the plastic yield point position indicates the critical point at which the cell wall structure begins to irreversibly destroy. The second wilting information is the state data of the leaf after the previous wilting treatment, and the change gradient of the leaf moisture content in the wilting degree index is determined by near-infrared spectroscopy or weight method, and is used to quantify the influence degree of water migration on the mechanical characteristic of the leaf during the wilting process.

[0164] The elastic deformation recovery rate is a dynamic parameter obtained by coupling analysis of the slope in the elastic deformation stage and the wilting degree index, and the implementation process is as follows: the attenuation function is established based on experimental data of the elastic modulus of the leaf under different wilting degrees, the slope offset reflects the actual influence of the current wilting state on the resilience of the leaf, and the normalization processing is used to eliminate the difference of the measurement system.

[0165] The cell wall rupture threshold value is a critical parameter obtained by correlating the plastic yield point position and the wilting degree index, the nonlinear regression model is constructed based on the statistical law of the yield point displacement of a large number of samples, the expected yield point position represents the structural strength benchmark under the ideal wilting state, and the position deviation value is converted into the rupture threshold value considering the constitutive relationship of material mechanics.

[0166] The typical rolling and twisting mode division in the mechanical characteristic coding library is based on the numerical distribution characteristics of the elastic deformation recovery rate, and the three characteristic intervals of light pressure shaping, medium pressure promoting breaking, and heavy pressure shaping correspond to different cell structure change mechanisms. The pressure action time length is determined by the cell wall rupture threshold and the process safety margin, and the corresponding relationship table between the characteristic interval and the rolling and twisting process parameters is established by considering the balance between production efficiency and quality formation. Preferably, the mechanical characteristic coding library can be calibrated according to the leaf characteristics of different tea tree varieties in actual application.

[0167] In this embodiment, the mechanical response characteristics are converted into quantifiable process control parameters through the dynamic correlation between the pressure-deformation characteristic parameters and the wilting degree index, wherein the elastic deformation recovery rate reflects the influence of the wilting degree on the plasticity of the leaves, and the cell wall rupture threshold represents the resistance limit of the leaf structure under a specific wilting state; the construction of the mechanical characteristic coding library realizes the accurate matching of the rolling and twisting process parameters and the leaf physical state, provides quantitative basis for standardized production, and ensures the balance between the optimal cell wall breaking rate and the shaping effect in the rolling and twisting process.

[0168] In some embodiments, according to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics, and the mechanical characteristic coding library, a mapping relationship network of different tea varieties and the optimal pressure gradient distribution and rolling and twisting rhythm mode is constructed, including:

[0169] The vein protrusion height distribution characteristics are spatially registered with the elastic deformation recovery rate to generate a regional pressure sensitivity distribution map;

[0170] The cell layer light transmission characteristics are associated with the cell wall rupture threshold for correlation analysis to establish a light transmission characteristic-pressure resistance strength correspondence table;

[0171] Based on the characteristic intervals in the mechanical characteristic coding library, the typical rolling and twisting mode parameter combinations of tea varieties are matched;

[0172] According to the regional pressure sensitivity distribution map, the light transmission characteristic-pressure resistance strength correspondence table, and the typical rolling and twisting mode parameter combinations, a mapping relationship network is constructed;

[0173] Based on the static physical parameters and the dynamic mechanical characteristics, a physical property conversion relationship function of the reference tea and the target tea is established, including:

[0174] The reference leaf edge thickness gradient characteristics of the reference tea and the target leaf edge thickness gradient characteristics of the target tea are extracted, and a preset thickness conversion coefficient is calculated;

[0175] A preset elastic deformation recovery rate is obtained according to the preset thickness conversion coefficient, and an elastic response correction factor is generated according to the preset elastic deformation recovery rate and the real-time elastic deformation recovery rate;

[0176] According to the preset thickness conversion coefficient and the elastic response correction factor, a physical property conversion relationship function is established.

[0177] In the embodiment, the static physical parameters are inherent morphological characteristics of tea leaves obtained by three-dimensional scanning, wherein the vein protrusion height distribution characteristics represent the spatial heterogeneity of the leaf surface structure and are used to indicate the sensitivity of different regions to external pressure; the regional pressure sensitivity distribution map is a two-dimensional heat map generated by spatial registration of the vein protrusion height distribution characteristics and the elastic deformation recovery rate, and the implementation process is as follows: the point cloud registration algorithm is used to align the geometric coordinates, and an interpolation calculation is performed to establish a continuous distribution model of the pressure sensitivity.

[0178] The light transmission property-pressure resistance corresponding relationship table is a quantitative control table established by correlating the light transmission property of the cell layer and the cell wall rupture threshold value, wherein the light transmission property of the cell layer reflects the cell arrangement density and the cell wall integrity, and the cell wall rupture threshold value represents the critical value of the leaf tissue resistance to mechanical damage. Preferably, the light transmission property-pressure resistance corresponding relationship table is constructed by using a piecewise linear regression method to ensure reasonable prediction accuracy of the pressure resistance in different light transmission property intervals.

[0179] The physical property conversion relationship function is a key mathematical model connecting the reference tea and the target tea, and in the establishment process thereof, the preset thickness conversion coefficient is calculated by the ratio of the reference leaf margin thickness gradient characteristics to the target leaf margin thickness gradient characteristics; the elastic response correction factor is generated by comparing the relative deviation of the preset elastic deformation recovery rate and the real-time measured value; and the final function form adopts a multivariate linear regression structure, which simultaneously considers the adjustment requirements of the thickness conversion and the elastic response in two dimensions.

[0180] In the embodiment, the intelligent matching network of the variety characteristics and the process parameters is constructed by multi-level feature fusion, wherein the regional pressure sensitivity distribution map solves the spatial optimization problem of pressure application, the light transmission property-pressure resistance corresponding relationship table realizes the correlation between the microstructure and the macro mechanical properties, and the physical property conversion relationship function guarantees the process adaptability adjustment of different raw materials. The mapping relationship network finally forms a closed-loop optimization system of "variety characteristics-process parameters", which provides decision support for personalized rolling and twisting process. In the embodiment, the accurate matching of the physical properties and the mechanical action in the tea processing process is realized through the collaborative analysis of the static characteristics and the dynamic response.

[0181] In some embodiments, according to the mapping relationship network and the physical property conversion relationship function, an initial adjustment strategy is generated and executed, and the initial adjustment strategy includes a pressure gradient distribution parameter, a rolling and twisting rhythm parameter, and a rolling and twisting time length parameter, which includes:

[0182] The initial pressure gradient distribution parameter is generated according to the regional pressure sensitivity distribution map, and the initial pressure gradient distribution parameter includes a first pressure value of the vein region and a second pressure value of the mesophyll region;

[0183] querying a pressure resistance threshold of the current tea variety from the light transmission characteristic-pressure resistance correspondence table;

[0184] setting a rolling rhythm parameter according to the pressure resistance threshold, the rolling rhythm parameter including a first duration of a pressure increasing phase and a second duration of a pressure decreasing phase;

[0185] determining a basic rolling duration parameter based on a typical rolling mode parameter combination;

[0186] correcting the basic rolling duration parameter through a physical property conversion relationship function to generate an actual rolling duration parameter adapted to the current tea leaf;

[0187] combining the initial pressure gradient distribution parameter, the rolling rhythm parameter and the actual rolling duration parameter to form an initial adjustment strategy;

[0188] controlling the rolling equipment to perform an initial rolling process according to the initial adjustment strategy.

[0189] In the embodiment, the initial adjustment strategy is a rolling process control scheme generated based on a mapping relationship network, wherein the pressure gradient distribution parameter is used to guide the rolling equipment to apply differential mechanical pressure to different leaf regions. Preferably, the first pressure value of the vein region is determined by the peak pressure in the region pressure sensitivity distribution map, and the second pressure value of the mesophyll region is calculated according to the sensitivity transition characteristics of adjacent regions. The pressure resistance threshold is a key safety parameter extracted from the light transmission characteristic-pressure resistance correspondence table, which is used to limit the maximum allowable operating pressure to prevent excessive damage to the cell structure.

[0190] The rolling rhythm parameter is a key variable for controlling the timing of pressure change, the first duration of the pressure increasing phase is dynamically adjusted according to the difference between the pressure resistance threshold and the current pressure, and the second duration of the pressure decreasing phase matches the elastic recovery characteristics of the leaf. The basic rolling duration parameter is derived from the typical rolling mode parameter combination in the mechanical feature coding library, and the actual rolling duration parameter is corrected by the physical property conversion relationship function, which can adapt to the thickness and elastic characteristics of the specific tea leaf raw material.

[0191] The execution process of the initial rolling process is as follows: the rolling equipment first sets the pressure partition according to the initial pressure gradient distribution parameter, then applies pressure cyclically according to the rolling rhythm parameter, and finally terminates the operation when the actual rolling duration parameter is reached. Preferably, the pressure application process adopts a gradual loading mode to avoid sudden stress that may cause leaf fracture; a moderate gap is maintained during the pressure decreasing phase to facilitate uniform exudation of cell contents.

[0192] The embodiment realizes precise control of the rolling process by combining static feature mapping with dynamic parameter adjustment. The pressure gradient distribution parameter ensures that the mechanical force matches the leaf structure characteristics, the rolling rhythm parameter maintains the balance between cell wall breaking and shape forming, and the actual rolling duration parameter ensures the adaptability of process intensity and material characteristics. The formation process of the initial adjustment strategy forms a closed-loop control from feature analysis to parameter generation and process execution, providing a standardized and customizable solution for tea rolling processing. The whole embodiment optimizes multiple parameters to maximize tea quality characteristics while ensuring processing efficiency.

[0193] In some embodiments, the real-time state information is compared with the predicted state information in the initial adjustment strategy, and the pressure gradient distribution parameter, the rolling rhythm parameter and the rolling duration parameter are dynamically adjusted according to the comparison result to generate and update the optimized adjustment strategy, including:

[0194] The current cell damage rate and the current red edge formation degree in the real-time state information are obtained;

[0195] The current cell damage rate is compared with the expected cell damage rate in the initial adjustment strategy to calculate a damage rate deviation value;

[0196] The current red edge formation degree is compared with the expected red edge formation degree in the initial adjustment strategy to calculate a red edge deviation value;

[0197] The pressure gradient distribution parameter is adjusted according to the damage rate deviation value, including:

[0198] When the damage rate deviation value is negative, the first pressure value and the second pressure value are increased by a preset gradient;

[0199] When the damage rate deviation value is positive, the first pressure value and the second pressure value are decreased by a preset proportion;

[0200] The rolling rhythm parameter is adjusted according to the red edge deviation value, including:

[0201] When the red edge deviation value is negative, the first duration is extended and the second duration is shortened;

[0202] When the red edge deviation value is positive, the first duration is shortened and the second duration is extended;

[0203] The actual rolling duration parameter is recalculated based on the adjusted pressure gradient distribution parameter and the adjusted rolling rhythm parameter;

[0204] The adjusted pressure gradient distribution parameter, the adjusted rolling rhythm parameter and the recalculated actual rolling duration parameter are combined to form the optimized adjustment strategy;

[0205] The control parameters of the rolling equipment are updated according to the optimized adjustment strategy.

[0206] In the present embodiment, the real-time state information is the dynamic parameters of the rolling and twisting process acquired by the online monitoring system, wherein the current cell damage rate is determined by microscopic image analysis technology, representing the degree of damage to the leaf cell structure; the current red edge formation degree is extracted by spectral reflection characteristics, reflecting the progress of tea oxidation and fermentation. The expected cell damage rate and the expected red edge formation degree are preset process target values in the initial adjustment strategy, determined by the optimal parameters in the historical process database. The damage rate deviation value and the red edge deviation value are obtained by the difference between the real-time measurement value and the expected value, used to quantify the deviation of the process execution effect from the target.

[0207] The adjustment of the pressure gradient distribution parameter follows the principle of gradual optimization. When the damage rate deviation value is negative, the first pressure value and the second pressure value are increased by a preset gradient value, which is set according to the leaf thickness and elastic modulus characteristics. When the deviation value is positive, the pressure value is decreased by a preset proportion, which takes into account the tolerance limit of the cell wall structure. The adjustment of the rolling and twisting rhythm parameter is based on the oxidation and fermentation kinetics principle. When the red edge deviation value is negative, the first duration of the pressure stage is lengthened to promote enzymatic reactions, while the second duration of the loose pressure stage is shortened. When the deviation value is positive, the opposite adjustment strategy is adopted. Further, the recalculation of the actual rolling and twisting duration parameter takes into account the cumulative effect of pressure intensity and rhythm changes, achieved by a time integration algorithm.

[0208] The formation process of the optimization adjustment strategy is as follows: first, adjust each sub-parameter independently according to the deviation analysis result, then verify the process feasibility through parameter coupling analysis, and finally generate coordinated control instructions. The update of the rolling and twisting equipment control parameters uses an incremental writing method to ensure smooth transition of process adjustment. Preferably, safety boundary conditions are set during parameter adjustment to prevent process from losing control due to continuous adjustment.

[0209] The present embodiment realizes the adaptive optimization of the rolling and twisting process through the establishment of a real-time feedback mechanism of "monitoring-comparison-adjustment". The cell damage rate deviation guides the precise regulation of mechanical force, and the red edge formation degree deviation adjusts the spatiotemporal distribution of oxidation and fermentation, both of which work together to ensure that the formation of tea quality is on the best trajectory. The dynamic generation of the optimization adjustment strategy embodies the intelligent features of data-driven decision-making, enabling the traditional rolling and twisting process to have self-correcting ability. The present embodiment converts empirical operation into quantifiable process management through closed-loop control, significantly improving the consistency and controllability of tea processing.

[0210] In the second aspect, the present embodiment also provides a rolling and twisting parameter dynamic adjustment system suitable for the method of the first aspect, the system comprising a collection module, a data processing module, and a logic operation module.

[0211] The collection module is configured to collect original tea leaf characteristic information, and the original tea leaf characteristic information includes first leaf image information, first response information, and first withering information.

[0212] The data processing module is configured to pre-process the original tea leaf characteristic information to obtain processed tea leaf characteristic information, and the processed tea leaf characteristic information includes second leaf image information after process alignment, second response information, and second withering information.

[0213] The logic operation module is configured to perform three-dimensional point cloud reconstruction on the second leaf image information, extract static physical parameters, and perform hyperspectral analysis on the second leaf image information to obtain cell layer light transmission characteristics, wherein the static physical parameters include vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics; perform information correlation of the second response information and the second withering information in the rolling and twisting stage, extract dynamic mechanical characteristics, and establish a mechanical characteristic code library containing typical rolling and twisting modes, wherein the dynamic mechanical characteristics include elastic deformation recovery rate and cell wall rupture threshold, and the typical rolling and twisting modes include light pressure shaping, medium pressure promoting breaking, and heavy pressure shaping; construct a mapping relationship network of different tea varieties and optimal pressure gradient distribution and rolling and twisting rhythm modes according to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics, and the mechanical characteristic code library; establish a physical property conversion relationship function of reference tea and target tea according to the static physical parameters and the dynamic mechanical characteristics; generate an initial adjustment strategy according to the mapping relationship network and the physical property conversion relationship function, and execute the initial adjustment strategy, wherein the initial adjustment strategy includes pressure gradient distribution parameters, rolling and twisting rhythm parameters, and rolling and twisting time length parameters; collect real-time state information in the rolling and twisting process, wherein the real-time state information includes cell damage rate information and red edge formation degree information; compare the real-time state information with predicted state information in the initial adjustment strategy, dynamically adjust the pressure gradient distribution parameters, the rolling and twisting rhythm parameters, and the rolling and twisting time length parameters according to the comparison result, and generate an optimized adjustment strategy and update the optimized adjustment strategy.

[0214] In the embodiment, the collection module refers to a composite collection unit composed of an optical sensor array and a mechanical response testing device, which is configured to synchronously acquire morphological characteristics and mechanical response characteristics of tea leaves. The first leaf image information is collected by a multi-spectral imaging system and contains leaf surface texture and internal structure characteristics; the first response information is measured by a micro pressure sensor array and records the deformation recovery curve of the leaf under different pressures; and the first withering information is obtained by near-infrared spectrum analysis and reflects the water distribution and cell activity state of the tea leaves. The data processing module is equipped with a special time alignment algorithm, which can perform time synchronization processing on image data, mechanical data, and spectral data with different collection frequencies, and ensure the spatiotemporal consistency of subsequent analysis.

[0215] The logic operation module is an operation unit loaded with a three-dimensional reconstruction algorithm and a hyperspectral analysis model. The three-dimensional point cloud reconstruction is achieved by using a structured light scanning technology, which can accurately restore the three-dimensional morphological characteristics of the leaf. The hyperspectral analysis is achieved by using a characteristic band extraction algorithm to analyze the light transmission characteristics of the cell layer. The mechanical characteristic encoding library is a process knowledge base constructed by using a machine learning method. The light pressure shaping mode is suitable for the initial kneading stage to maintain the integrity of the leaf shape. The medium pressure promotes the breaking mode to accelerate the release of the cell content. The heavy pressure shaping mode realizes the final shaping of the tea leaf shape. The mapping relationship network adopts a graph neural network architecture, which can learn the nonlinear relationship between the tea characteristics and the process parameters.

[0216] The system establishes a digital mapping from the raw material characteristics to the process parameters through multi-source information collection and intelligent analysis, and realizes dynamic optimization based on real-time feedback. The acquisition module obtains the basic characteristic data of the tea leaf. The data processing module performs information fusion and standardization processing. The logic operation module completes the whole process calculation of feature extraction, knowledge base construction and strategy generation. During the rolling process, the system continuously monitors the process effect, automatically adjusts the control parameters through comparison and analysis, and forms a closed-loop optimization mechanism.

[0217] The embodiment realizes intelligent regulation and control of the tea rolling process by constructing a complete perception-decision-execution technical chain. The system can accurately identify the differences in raw material characteristics, automatically match the optimal process parameters, and real-time correct deviations during the processing process, which significantly improves the uniformity and quality stability of the tea rolling. The scheme realizes precise control and dynamic optimization of the process parameters in a data-driven manner, which provides reliable technical support for the standardization and intelligentization of tea processing.

[0218] In a third aspect, the embodiment also provides a computer readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the method of the first aspect.

[0219] The computer program involved in the embodiment can be stored in a computer device readable storage medium, including but not limited to magnetic disk, magnetic tape, magnetic card, floppy disk, flash memory, optical disc, optical card, read-only memory (ROM), random access memory (RAM), erasable programmable ROM (EPROM) and electrically erasable programmable ROM (EEPROM) and the like, and also includes other biological, physical or chemical structures that can realize similar or equivalent functions as the above-mentioned storage media, such as DNA, RNA, protein and the like units with information storage ability. In specific embodiments, the storage medium involved can be one of the above-mentioned medium types, or a combination of the above-mentioned medium types. In different embodiments, the computer program involved in the embodiment can be centrally stored in a single medium, or can be distributedly stored in multiple media. The storage medium containing the computer device readable storage medium can be a non-volatile memory or a random access memory. These computer device readable storage media can be built-in in the device, or connected with the device as an external device or part of the external device. In some embodiments, the storage medium with the computer device readable storage medium is deployed locally; in other embodiments, the storage medium can also be deployed remotely from the processor, for example, network attached storage accessed via RF circuit or external port and communication network, wherein the communication network can be Internet, one or more intranets, local area network (LAN), wide area network (WAN), storage area network (SAN) and the like, or appropriate combination thereof, as long as the access of the computer device to the storage medium can be realized. In addition, the computer program involved in the embodiment can be stored in plaintext / ciphertext form, or can be designed as training data, and integrated and reorganized by model training to be implicitly saved in the parameter state of the deep neural network or other machine learning model.

[0220] Please refer to Figure 3 In the fourth aspect, the embodiment also provides an electronic device 1, including a memory 11 and a processor 12, the memory 11 is used for storing one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor 12 to realize the method in the first aspect.

[0221] The processor described in the embodiment can be implemented by hardware, firmware, software or a combination thereof, and can use at least one of a circuit, a single or multiple application specific integrated circuits (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, a microprocessor, and other physical, biological or chemical structures that can realize the same or equivalent functions as the above-mentioned processors, such as biological neurons, quantum computing units, DNA computing units, etc., so that the processor can execute part or all of the steps or any combination of the steps mentioned in the computer programs or methods of various embodiments of the present application.

[0222] By adopting the above technical solutions, the present application is different from the prior art and has the following beneficial effects: the original tea characteristic information of tea leaves is collected through a multi-modal sensing fusion technology, static physical parameters and cell layer light transmission characteristics are extracted through three-dimensional point cloud reconstruction and hyperspectral analysis, a mechanical characteristic code library containing typical rolling modes is established in combination with dynamic mechanical characteristics, and a mapping relationship network of tea varieties and processing parameters is constructed. A physical property conversion relationship function is established based on the static physical parameters and the dynamic mechanical characteristics, process adaptation of different batches of raw materials is realized, through dynamic comparison of real-time state information and predicted state, the pressure gradient distribution parameters, the rolling rhythm parameters and the rolling time length parameters are adaptively adjusted, and a closed-loop optimization mechanism is formed. The above technical solutions convert traditional empirical operation into data-driven precise control, significantly improve the uniformity and quality stability of tea rolling while ensuring processing efficiency, and provide a standardized and intelligent solution for tea processing.

[0223] The above only describes some embodiments of the present application, and does not limit the protection scope of the present application, and any equivalent device or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for dynamically adjusting twisting parameters, characterized in that, The method comprises the following steps: Collecting original tea leaf characteristic information, the original tea leaf characteristic information comprising first leaf image information, first response information and first withering information; Pretreating the original tea leaf characteristic information to obtain processed tea leaf characteristic information, the processed tea leaf characteristic information comprising second leaf image information, second response information and second withering information after process alignment; Reconstructing the second leaf image information in three-dimensional point cloud, extracting static physical parameters, the static physical parameters comprising leaf vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics, and performing hyperspectral analysis on the second leaf image information to obtain cell layer light transmission characteristics; Correlating the second response information and the second withering information to extract dynamic mechanical characteristics, the dynamic mechanical characteristics comprising elastic deformation recovery rate and cell wall rupture threshold, and establishing a mechanical characteristic code library comprising typical rolling modes, the typical rolling modes comprising light pressure shaping, medium pressure promoting rupture and heavy pressure shaping; According to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics and the mechanical characteristic code library, a mapping relationship network of different tea varieties and optimal pressure gradient distribution and rolling rhythm mode is constructed; Based on the static physical parameters and the dynamic mechanical characteristics, a physical property conversion relationship function of reference tea and target tea is established; According to the mapping relationship network and the physical property conversion relationship function, an initial adjustment strategy is generated and executed, the initial adjustment strategy comprising pressure gradient distribution parameters, rolling rhythm parameters and rolling time length parameters; Collecting real-time state information in the rolling process, the real-time state information comprising cell damage rate information and red edge formation degree information; Comparing the real-time state information with the predicted state information in the initial adjustment strategy, dynamically adjusting the pressure gradient distribution parameters, the rolling rhythm parameters and the rolling time length parameters according to the comparison result, generating an optimized adjustment strategy and updating.

2. The method of claim 1, wherein, Reconstructing the second leaf image information in three-dimensional point cloud, extracting static physical parameters, the static physical parameters comprising leaf vein protrusion height distribution characteristics and leaf edge thickness gradient characteristics, comprising: Obtaining multi-view point cloud information of the leaf surface, the multi-view point cloud information comprising overhead angle point cloud, side view angle point cloud and oblique view angle point cloud; Performing registration fusion processing on the multi-view point cloud information, the registration fusion processing comprising unifying the overhead angle point cloud, the side view angle point cloud and the oblique view angle point cloud to the same coordinate system by using the iterative closest point algorithm, and realizing point cloud density equalization by voxel grid filtering to obtain registered point cloud data; Reconstructing a leaf surface three-dimensional geometric model based on the registered point cloud data, the leaf surface three-dimensional geometric model being configured to generate a closed surface by using a Poisson surface reconstruction algorithm, and applying Laplace smoothing to eliminate surface noise; In the leaf surface three-dimensional geometric model, calculating point cloud Gaussian curvature distribution, extracting curvature maximum points to constitute a leaf vein main ridge line contour, and obtaining a continuous ridge line trajectory by B-spline curve fitting to obtain a leaf vein main ridge line; Establishing a local coordinate system in the normal plane of the ridge line trajectory of the leaf vein main ridge line, measuring the vertical distance of each sampling point to the reference surface, generating a height-position distribution function, and extracting a protrusion height distribution curve, denoted as leaf vein protrusion height information; Setting equidistant measuring points along the blade edge contour line in the three-dimensional geometric model of the blade surface, obtaining the blade thickness at each point by the ray method, and calculating the thickness gradient between adjacent points to obtain the thickness gradient change rate, denoted as the blade edge thickness information; Establishing a spatial correlation matrix of the vein protrusion height information and the blade edge thickness information, including: Aligning the vein height distribution curve and the blade edge thickness gradient sequence in time and space; Constructing a two-dimensional feature mapping table reflecting the correlation between the two to obtain the spatial correlation matrix; Extracting the static physical parameters in the spatial correlation matrix.

3. The twisting parameter dynamic adjustment method according to claim 2, characterized in that, Performing hyperspectral analysis on the second leaf image information to obtain the cell layer light transmission characteristics, including: Extracting the characteristic spectral curve of the mesophyll region in the second leaf image information; Selecting the first spectral reflectance of the 1450nm water molecule absorption characteristic band and the second spectral reflectance of the 1940nm water molecule absorption characteristic band in the characteristic spectral curve; Calculating the absorption depth ratio of the first spectral reflectance and the second spectral reflectance to obtain the light transmission characteristic conversion coefficient; Spatially registering the characteristic spectral curve with the three-dimensional geometric model of the blade surface; Establishing a light transmission rate calculation model according to the light transmission characteristic conversion coefficient, and converting the registered three-dimensional geometric model into a cell layer light transmission characteristic spatial distribution map; Extracting the cell layer light transmission characteristics from the cell layer light transmission characteristic spatial distribution map.

4. The method of claim 1, wherein, Correlating the second response information and the second wilting information to extract dynamic mechanical characteristics, including: Analyzing the pressure-deformation characteristic parameters from the second response information, including the elastic deformation stage slope and the plastic yield point position; Extracting the wilting degree index from the second wilting information, including the leaf water content change gradient; Coupling the elastic deformation stage slope with the wilting degree index for coupling analysis to generate an elastic deformation recovery rate, including: Establishing a decay function of the elastic deformation stage slope with the wilting degree; Calculating the slope offset at the current wilting degree in the decay function; Normalizing the slope offset to generate the elastic deformation recovery rate; Correlating the plastic yield point position with the wilting degree index to obtain the cell wall rupture threshold, including: Constructing a nonlinear regression model of the yield point displacement and the wilting degree; Interpolating the current wilting degree in the nonlinear regression model to obtain the expected yield point position; Calculating the deviation value between the measured yield point position and the expected yield point position to obtain the position deviation value; Converting the position deviation value into the cell wall rupture threshold; And, establishing a mechanical characteristic coding library containing typical rolling patterns, including: Dividing the elastic deformation recovery rate into three characteristic intervals: light pressure shaping, medium pressure promoting breaking, and heavy pressure shaping; Setting the pressure action time length of each characteristic interval based on the cell wall rupture threshold; Constructing a correspondence table of the characteristic intervals and rolling process parameters based on the pressure action time length to obtain the mechanical characteristic coding library.

5. The method of claim 1, wherein, According to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics, and the mechanical characteristic coding library, a mapping relationship network of different tea varieties and the optimal pressure gradient distribution and rolling rhythm mode is constructed, including: The leaf vein protrusion height distribution characteristics are spatially registered with the elastic deformation recovery rate to generate a regional pressure sensitivity distribution map; The cell layer light transmission characteristics are associated with the cell wall rupture threshold to establish a light transmission characteristic-pressure resistance strength correspondence table; Based on the feature interval in the mechanical characteristic coding library, the typical rolling mode parameter combination of the tea variety is matched; According to the regional pressure sensitivity distribution map, the light transmission characteristic-pressure resistance strength correspondence table, and the typical rolling mode parameter combination, a mapping relationship network is constructed; Based on the static physical parameters and the dynamic mechanical characteristics, a physical property conversion relationship function of the reference tea and the target tea is established, including: The reference leaf margin thickness gradient characteristics of the reference tea and the target leaf margin thickness gradient characteristics of the target tea are extracted, and a preset thickness conversion coefficient is calculated; A preset elastic deformation recovery rate is obtained according to the preset thickness conversion coefficient, and an elastic response correction factor is generated according to the preset elastic deformation recovery rate and the real-time elastic deformation recovery rate; The physical property conversion relationship function is established according to the preset thickness conversion coefficient and the elastic response correction factor.

6. The twisting parameter dynamic adjustment method according to claim 5, characterized in that, According to the mapping relationship network and the physical property conversion relationship function, an initial adjustment strategy is generated and executed, and the initial adjustment strategy includes a pressure gradient distribution parameter, a rolling rhythm parameter, and a rolling time length parameter, including: An initial pressure gradient distribution parameter is generated according to the regional pressure sensitivity distribution map, and the initial pressure gradient distribution parameter includes a first pressure value of the vein area and a second pressure value of the mesophyll area; The pressure resistance strength threshold of the current tea variety is queried from the light transmission characteristic-pressure resistance strength correspondence table; The rolling rhythm parameter is set according to the pressure resistance strength threshold, and the rolling rhythm parameter includes a first duration of the pressure increasing stage and a second duration of the pressure releasing stage; The basic rolling time length parameter is determined based on the typical rolling mode parameter combination; The basic rolling time length parameter is corrected by the physical property conversion relationship function to generate an actual rolling time length parameter adapted to the current tea; The initial pressure gradient distribution parameter, the rolling rhythm parameter, and the actual rolling time length parameter are combined to form an initial adjustment strategy; The initial rolling process is controlled by the rolling equipment according to the initial adjustment strategy.

7. The twisting parameter dynamic adjustment method according to claim 6, characterized in that, The real-time state information is compared with the predicted state information in the initial adjustment strategy, and the pressure gradient distribution parameter, the rolling rhythm parameter, and the rolling time length parameter are dynamically adjusted according to the comparison result to generate an optimized adjustment strategy and update, including: The current cell damage rate and the current red edge formation degree in the real-time state information are obtained; The current cell damage rate is compared with the expected cell damage rate in the initial adjustment strategy to calculate a damage rate deviation value; The current red edge formation degree is compared with the expected red edge formation degree in the initial adjustment strategy to calculate a red edge deviation value; The pressure gradient distribution parameter is adjusted according to the damage rate deviation value, including: when the breakage rate deviation value is negative, increasing the first pressure value and the second pressure value by a preset gradient; when the breakage rate deviation value is positive, decreasing the first pressure value and the second pressure value by a preset ratio; adjusting the rolling rhythm parameters according to the red edge deviation value, including: when the red edge deviation value is negative, lengthening the first duration and shortening the second duration; when the red edge deviation value is positive, shortening the first duration and lengthening the second duration; recomputing the actual rolling time length parameters based on the adjusted pressure gradient distribution parameters and the adjusted rolling rhythm parameters; combining the adjusted pressure gradient distribution parameters, the adjusted rolling rhythm parameters and the recomputed actual rolling time length parameters to form an optimized adjustment strategy; updating the control parameters of the rolling equipment according to the optimized adjustment strategy.

8. A twisting parameter dynamic adjustment system, characterized in that, The system is suitable for the method of any one of claims 1 to 7, and comprises a collection module, a data processing module and a logical operation module; The collection module is used for collecting original tea leaf characteristic information, and the original tea leaf characteristic information includes first leaf image information, first response information and first withering information. The data processing module is used for preprocessing the original tea leaf characteristic information to obtain processed tea leaf characteristic information, and the processed tea leaf characteristic information includes second leaf image information, second response information and second withering information after process alignment. The logical operation module is used for performing three-dimensional point cloud reconstruction on the second leaf image information, extracting static physical parameters, and performing hyperspectral analysis on the second leaf image information to obtain cell layer light transmission characteristics; performing rolling stage information correlation on the second response information and the second withering information, extracting dynamic mechanical characteristics, and establishing a mechanical characteristic coding library containing typical rolling modes; the typical rolling modes include light pressure shaping, medium pressure promoting breaking and heavy pressure shaping; constructing a mapping relationship network of different tea varieties and optimal pressure gradient distribution and rolling rhythm mode according to the static physical parameters, the cell layer light transmission characteristics, the dynamic mechanical characteristics and the mechanical characteristic coding library; establishing a physical property conversion relationship function of reference tea and target tea based on the static physical parameters and the dynamic mechanical characteristics; generating an initial adjustment strategy and executing the initial adjustment strategy according to the mapping relationship network and the physical property conversion relationship function, and the initial adjustment strategy includes pressure gradient distribution parameters, rolling rhythm parameters and rolling time length parameters; collecting real-time state information in the rolling process, and the real-time state information includes cell breakage rate information and red edge formation degree information; comparing the real-time state information with the predicted state information in the initial adjustment strategy, dynamically adjusting the pressure gradient distribution parameters, the rolling rhythm parameters and the rolling time length parameters according to the comparison result, generating an optimized adjustment strategy and updating.

9. A computer readable storage medium having stored thereon computer program instructions, wherein, The computer program instructions, when executed by the processor, implement the method of any one of claims 1 to 7.

10. An electronic device comprising a memory and a processor, characterized in that The memory is configured to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method according to any one of claims 1 to 7.

Images