Tea leaf expert remote consultation method based on artificial intelligence

Abstract

The invention relates to the technical field of tea intelligent consultation, and discloses a tea expert remote consultation method based on artificial intelligence. The method comprises the following steps: processing growth environment data through a tea tree physiological status dynamic analysis module to generate a tea tree physiological status feature vector; historical agronomic operation nodes are extracted based on management records, and the physiological state feature vectors are associated to construct a tea tree response mode knowledge base; triggering a consultation request according to tea tree phenotypic data collected in real time, and integrating the physiological status feature vector and the response mode knowledge base through a cross-dimension feature fusion engine to generate a multi-factor association decision tree; performing iterative adjustment on the node weight of the decision tree by adopting a self-adaptive decision optimization module, and outputting a tea tree growth abnormality diagnosis result and an agricultural intervention scheme; and transmitting the scheme to terminal equipment through a remote consultation interface, and updating an operation validity mark of the response mode knowledge base. The method realizes remote precise consultation of tea planting and assists scientific management of a tea garden.

Description

Technical Field

[0001] This invention relates to the field of intelligent tea consultation technology, specifically to a remote consultation method for tea experts based on artificial intelligence. Background Technology

[0002] In the tea cultivation industry, precise control and scientific management of tea tree growth are crucial for ensuring tea quality and yield. In traditional tea cultivation, the diagnosis of abnormal tea tree growth and the development of management plans heavily rely on the on-site investigations and experience of professional technicians. However, major tea-producing areas are often located in mountainous and hilly regions with complex terrain. The scattered locations of tea gardens and inconvenient transportation make it difficult for professional experts to conduct regular on-site visits, leaving many small and medium-sized tea gardens facing a "nowhere to turn for advice" dilemma. When tea trees exhibit problems such as slow growth, pest and disease infestation, and declining quality, growers often can only attempt to address the issues based on their limited experience or seek help from distant experts through simple means such as phone calls or pictures. This makes it difficult to comprehensively and accurately convey information about tea tree growth, leading to frequent diagnostic errors and inappropriate interventions.

[0003] Tea tree growth is influenced by a synergistic effect of multiple environmental factors. Soil pH, nutrient content, and organic matter ratio directly affect root absorption. Temporal variations in meteorological factors such as temperature, precipitation, light, and humidity determine photosynthetic efficiency and the rate of nutrient accumulation. Biological stress signals, such as pests and diseases, directly threaten tea tree health. In traditional consultation models, monitoring of these environmental factors relies heavily on manual sampling and simple instrument measurements. Data collection is discontinuous and incomplete, making it difficult to capture the dynamic correlation between environmental changes and tea tree growth. Furthermore, tea garden management records are mostly in the form of paper ledgers or scattered electronic documents, lacking standardized organization. The causal relationship between historical agronomical measures (such as fertilization, pruning, and pest and disease control) and tea tree growth responses is difficult to trace, resulting in the inability to effectively preserve and reuse valuable planting experience.

[0004] Existing remote consultation methods generally suffer from a lack of information dimensions. Relying solely on grower descriptions or partial image information, experts cannot comprehensively grasp key data such as tea garden soil, weather, biological stress, and historical management. Decision-making is easily limited by subjective experience, making it difficult to formulate precise and effective intervention plans. Furthermore, due to the lack of a dynamically updated knowledge system, the differences in growth characteristics of different regions and tea varieties cannot be fully considered, resulting in insufficient universality and specificity of consultation plans, making it difficult to adapt to the diverse needs of tea garden cultivation. With the large-scale and refined development of the tea industry, traditional consultation models relying on manual experience and limited information can no longer meet the needs of modern scientific tea garden management. There is an urgent need for a remote consultation method that can integrate multi-source data, achieve intelligent analysis, and overcome the limitations of time and space. Summary of the Invention

[0005] The purpose of this invention is to provide a remote consultation method for tea experts based on artificial intelligence, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a remote consultation method for tea experts based on artificial intelligence, the method comprising:

[0007] Acquire tea tree growth environment data and tea garden management operation records. The tea tree growth environment data includes soil physicochemical indicators, temporal changes of meteorological factors, and biological stress signals.

[0008] The tea tree physiological state dynamic analysis module processes the tea tree growth environment data to generate a tea tree physiological state feature vector. Based on the tea garden management operation records, historical agronomic operation nodes are extracted and associated with the tea tree physiological state feature vector to construct a tea tree response pattern knowledge base. Consultation requests are triggered based on real-time collected tea tree phenotypic data. A cross-dimensional feature fusion engine integrates the tea tree physiological state feature vector with the tea tree response pattern knowledge base to generate a multi-factor association decision tree. An adaptive decision optimization module iteratively adjusts the node weights of the multi-factor association decision tree, outputting tea tree growth anomaly diagnosis results and agronomic intervention plans. The agronomic intervention plans are transmitted to the terminal device via a remote consultation interface, while simultaneously updating the operation validity markers in the tea tree response pattern knowledge base.

[0009] Preferably, the step of processing the tea tree growth environment data through the tea tree physiological state dynamic analysis module to generate a tea tree physiological state feature vector includes: separating the time-series data of nitrogen, phosphorus, and potassium content and the spatial distribution data of trace elements from the soil physicochemical indicators, and inputting them into a soil nutrient response model to generate soil fertility characteristics; analyzing the light intensity fluctuation pattern and temperature and humidity co-change curve in the time-series changes of meteorological factors, and outputting the meteorological stress index through a microclimate impact assessment model; identifying the spectrum of pest and disease characteristics and the activity trajectory of natural enemies in the biological stress signals, and generating ecological balance parameters using a biological interaction relationship map; and fusing the soil fertility characteristics, meteorological stress index, and ecological balance parameters, and generating a tea tree physiological state feature vector through multi-dimensional feature dimensionality reduction processing.

[0010] Preferably, the step of constructing a tea tree response pattern knowledge base by associating the tea tree physiological state feature vector includes: matching pruning time markers and fertilizer application change records in the historical agronomic operation nodes to extract the time series sequence of agronomic operation intensity; establishing a dynamic response relationship between the time series sequence of agronomic operation intensity and the tea tree physiological state feature vector to generate an operation response lag period matrix; processing the operation response lag period matrix through a tea tree variety adaptability analysis model to classify and store agronomic operation threshold parameters for different tea tree varieties; and integrating the operation response lag period matrix and the agronomic operation threshold parameters to construct a tea tree response pattern knowledge base containing tea tree growth stage markers.

[0011] Preferably, the step of integrating the tea tree physiological state feature vector and the tea tree response pattern knowledge base through a cross-dimensional feature fusion engine to generate a multi-factor association decision tree includes: extracting bud and leaf color distribution parameters and branch elongation rate from the current tea tree phenotypic data to generate phenotypic anomaly feature codes; spatiotemporally aligning the phenotypic anomaly feature codes with the tea tree physiological state feature vector, and generating environmental stress association factors through a feature cross-validation module; matching the growth stage markers of the same tea tree variety in the tea tree response pattern knowledge base, and calling the corresponding agronomic operation threshold parameters; and fusing the environmental stress association factors and the agronomic operation threshold parameters to construct a multi-factor association decision tree with tea tree organs as nodes.

[0012] Preferably, the step of using an adaptive decision optimization module to iteratively adjust the node weights of the multi-factor association decision tree includes: identifying the strength of the association path between leaf nodes and root nodes in the multi-factor association decision tree, and generating organ interaction weight coefficients; correcting the initial assignment of the organ interaction weight coefficients based on intervention effect feedback data in a historical consultation case library; adjusting the decision priority of different organ nodes through a tea tree growth stage adaptive function, and generating a dynamic weight allocation table; and reorganizing the nodes of the multi-factor association decision tree according to the dynamic weight allocation table, and outputting a tea tree growth abnormality diagnosis result containing time-sensitive markers.

[0013] Preferably, the output of the tea tree growth abnormality diagnosis results and agronomic intervention plan includes: parsing the stress type code and severity level parameter in the tea tree growth abnormality diagnosis results to generate abnormality classification labels; associating historical intervention records with the same abnormality classification labels in the tea tree response pattern knowledge base to extract agronomic operation combination sequences; processing the agronomic operation combination sequences through a timeliness constraint model to screen the executable operation set that conforms to the current phenological period; integrating weather forecast data and real-time soil moisture monitoring values ​​to optimize the implementation time window of the executable operation set and generate an agronomic intervention plan that includes dosage adjustment parameters.

[0014] Preferably, transmitting the agronomic intervention plan to the terminal device via the remote consultation interface includes: decomposing the operation steps in the agronomic intervention plan into equipment control instructions and manual operation guidance text; adapting the equipment control instructions to the communication protocol of the intelligent irrigation equipment and fertilizer machinery through a protocol conversion gateway; synthesizing voice prompts and three-dimensional motion demonstration data of the manual operation guidance text using a multimodal interaction engine; and encapsulating the equipment control instructions, the voice prompts, and the three-dimensional motion demonstration data to form a consultation response data packet for transmission to the terminal device.

[0015] Preferably, updating the operation validity marker of the tea tree response pattern knowledge base includes: collecting agronomic intervention execution logs fed back by terminal devices, and analyzing actual operation time deviation and dosage deviation data; comparing the change rate of the tea tree phenotypic data before and after intervention to calculate the operation response efficiency coefficient; associating the original agronomic operation threshold parameters in the tea tree response pattern knowledge base to correct the operation trigger condition threshold; binding the operation response efficiency coefficient with the corrected operation trigger condition threshold, and updating the operation validity marker of the tea tree response pattern knowledge base.

[0016] Preferably, before acquiring tea tree growth environment data and tea garden management operation records, the method further includes: deploying multispectral sensing equipment to collect tea tree canopy reflectance data and generating a spatial distribution map of chlorophyll content; installing underground rhizosphere monitoring probes to acquire time-series data of root exudate concentration and constructing root zone microbial activity indicators; and using a wireless sensor network to aggregate the spatial distribution map of chlorophyll content and the root zone microbial activity indicators, and fusing topographic relief parameters from satellite remote sensing images to generate a basic dataset of tea tree growth environment.

[0017] Preferably, the step of triggering a consultation request based on real-time collected tea tree phenotypic data includes: capturing the sequence of changes in the growth angle of tea tree shoots using a depth camera device to generate a morphogenesis dynamic curve; measuring the weight increment of fresh leaves using a high-precision weighing sensor and generating a substance accumulation rate parameter in combination with the harvesting timestamp; and automatically generating a consultation request containing a geolocation code when the morphogenesis dynamic curve deviates from the variety benchmark model or the substance accumulation rate parameter is lower than a threshold.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0019] This method provides rich and systematic basic information support for remote consultation by tea experts by comprehensively collecting data on tea tree growth environment and tea garden management operation records. The growth environment data covers soil physicochemical indicators, temporal changes in meteorological factors, and biological stress signals, breaking through the limitations of fragmented environmental information in traditional consultations. This allows for a complete presentation of external influencing factors on tea tree growth, helping to more accurately trace the root causes of changes in tea tree growth status. The inclusion of tea garden management operation records links historical agronomic measures with tea tree growth responses, avoiding the drawbacks of viewing current problems in isolation and enabling the consultation process to form more targeted judgments based on past management experience.

[0020] The dynamic analysis module for tea tree physiological states generates physiological state feature vectors, transforming the complex growth state of tea trees into quantifiable and analyzable feature data. This changes the vague state of traditional consultations that rely on visual observation and experience-based descriptions. This quantitative presentation allows remote experts or intelligent systems to more intuitively grasp changes in the physiological functions of tea trees, such as nutrient absorption, metabolic levels, and stress resistance. This provides clear analytical basis for subsequent abnormal diagnosis, making the diagnostic process more scientific and objective, and reducing biases caused by subjective judgment.

[0021] A tea tree response pattern knowledge base, constructed based on historical agronomic operation nodes and physiological state feature vectors, enables the systematic accumulation and reuse of tea garden management experience. This knowledge base is not a static collection of experience, but a dynamic system continuously enriched and improved by linking new management records and physiological state data. This allows for the effective accumulation of response patterns of different regions and tea varieties to various agronomic measures under different environmental conditions. Subsequent consultations can directly access relevant historical data and response patterns, avoiding redundant trial and error and wasting experience, enabling new consultation requests to form more rational decisions based on past experience.

[0022] The mechanism that triggers consultation requests based on real-time collected tea tree phenotypic data ensures timely consultation responses, changing the situation of delayed problem discovery and untimely intervention in traditional consultations. When tea trees exhibit phenotypic abnormalities such as yellowing leaves and stunted growth, the consultation process can be quickly initiated, combining real-time environmental data and historical knowledge for analysis, avoiding irreparable losses once the problem worsens. The cross-dimensional feature fusion engine integrates physiological state feature vectors and response pattern knowledge bases to generate multi-factor association decision trees, organically combining multi-dimensional information such as environmental factors, physiological states, and historical experience. This overcomes the limitations of single-source information analysis, allowing the decision-making process to comprehensively consider various influencing factors and form more comprehensive diagnostic results and intervention plans.

[0023] The adaptive decision optimization module iteratively adjusts the node weights of the multi-factor association decision tree, enabling the decision-making process to adapt to the unique characteristics of different tea gardens and their dynamically changing growth environments. Different tea gardens have varying soil characteristics, climate conditions, and tea varieties, and the needs of the same tea garden also change at different growth stages. The iterative weight adjustment mechanism allows the decision tree to continuously learn and adapt to these differences, making the output agronomic intervention plan more tailored to the specific conditions of the tea garden, thus improving the operability and effectiveness of the plan.

[0024] The remote consultation interface enables the rapid transmission of agronomic intervention plans, breaking geographical limitations and allowing tea gardens in remote areas to receive timely professional consultation support. This solves the problems of high cost and limited coverage associated with traditional on-site expert consultations. Simultaneously, the real-time updating of the tea tree response pattern knowledge base through operational effectiveness tagging forms a closed-loop mechanism of "consultation-intervention-feedback-optimization," allowing the knowledge base to continuously absorb new practical experience, constantly improving the accuracy and reliability of subsequent consultations, and driving the continuous evolution of tea experts' remote consultation capabilities. Attached Figure Description

[0025] Figure 1 This is a schematic diagram illustrating the working principle of the artificial intelligence-based remote consultation method for tea experts according to the present invention.

[0026] Figure 2 A flowchart for generating feature vectors of tea tree physiological states;

[0027] Figure 3 A flowchart for generating a multi-factor association decision tree;

[0028] Figure 4 This is a flowchart showing the diagnostic results of abnormal tea tree growth and the output of agronomic intervention plans. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Please see Figure 1 This invention provides a remote consultation method for tea experts based on artificial intelligence, the method comprising:

[0031] The system first deploys equipment such as multispectral sensors, underground rhizosphere monitoring probes, and weather stations to collect real-time data on the tea tree's growth environment, including soil physicochemical indicators, temporal changes in meteorological factors, and biological stress signals. Simultaneously, it records historical data on tea garden management operations, forming a structured database. A dynamic analysis module for tea tree physiological status extracts and fuses features from the raw data, generating feature vectors representing the health status of the tea trees. The system establishes a knowledge base of tea tree response patterns, storing the dynamic correlation between different agronomic operations and the physiological state of tea trees. When abnormal tea tree phenotypic data is detected, a cross-dimensional feature fusion engine matches and analyzes real-time data with the knowledge base, constructing a multi-factor association decision tree. An adaptive decision optimization module continuously adjusts the weights of decision nodes based on historical cases, ultimately outputting accurate diagnostic results and agronomic intervention plans. A remote consultation interface converts the plans into equipment control commands and visual operation guides, synchronously updating the operation validity markers in the knowledge base.

[0032] Example 1: See Figure 2 The dynamic analysis module for tea tree physiological status employs a multi-level data processing workflow to achieve in-depth analysis and feature extraction of tea tree growth environment data. This module first receives raw monitoring data from various sensors, including soil physicochemical indicators, temporal changes in meteorological factors, and biological stress signals. The soil data processing unit is equipped with high-precision ion-selective electrodes to collect parameters such as soil conductivity, pH value, and redox potential in real time, and obtains the content distribution of macroelements such as nitrogen, phosphorus, and potassium, as well as microelements such as iron, manganese, and zinc, through spectral analysis technology. Timestamps are used during data acquisition to ensure the continuity of time-series data. The soil nutrient response model, trained based on historical data, can predict nutrient consumption trends over a future period based on current soil conditions and generate a soil fertility feature vector containing key indicators such as nutrient availability, buffering capacity, and microbial activity.

[0033] The meteorological data processing unit integrates multi-source meteorological sensors, including a light intensity meter, temperature and humidity sensors, an anemometer, and a rainfall monitoring device. Light intensity data is processed using a shading compensation algorithm to eliminate measurement errors caused by tea canopy shading, generating a continuous photosynthetically active radiation curve. Temperature and humidity data are analyzed through a collaborative analysis module to calculate dew point temperature and evapotranspiration, assessing the impact of current microclimate conditions on tea tree water use efficiency. The meteorological stress index calculation model comprehensively considers factors such as extreme high and low temperatures, drought, and strong sunlight, outputting a comprehensive score to quantify the potential stress of meteorological conditions on tea tree growth.

[0034] The biological stress signal processing unit employs a combination of machine vision and acoustic monitoring to identify pest and disease characteristics and the activity of natural enemies. High-resolution cameras regularly capture images of tea leaves and stems, and convolutional neural networks are used to analyze lesion morphology, pest feeding marks, and pathogen infection characteristics. Simultaneously, acoustic sensors deployed within the tea garden capture insect wingbeat frequencies and the acoustic signatures of natural enemies, combined with infrared cameras to record the diurnal activity patterns of predatory insects. The biological interaction map, constructed based on an ecological model, can predict the ecological balance over a future period based on current pest and disease population densities and the number of natural enemies, and output ecological balance parameters to assess the potential risk of biological stress.

[0035] The feature fusion stage employs dimensionality reduction techniques to transform multidimensional environmental data into a compact and representative feature vector representing the physiological state of tea trees. Principal component analysis (PCA) is used to extract the main directions of variation in data such as soil fertility, meteorological stress, and ecological balance, generating a 12-dimensional feature vector. Each dimension corresponds to a state indicator of a key physiological function of the tea tree. For example, the first principal component may reflect the overall supply of water and nutrients, the second principal component characterizes the synergistic effect of light and temperature, and the third principal component relates to pest and disease pressure and ecological regulation capacity. This feature vector serves as the foundational input for subsequent decision analysis, comprehensively reflecting the real-time growth status of the tea trees.

[0036] The construction of the tea tree response pattern knowledge base employs time-series alignment and dynamic response modeling techniques. The system extracts historical agronomic operation nodes from tea garden management records, including key events such as pruning, fertilization, irrigation, and pest and disease control, recording the specific time, dosage, and implementation method of each operation. Pruning operations are quantified using 3D laser scanning, recording changes in cut height, branch removal ratio, and canopy light transmittance. Fertilization data includes fertilizer type, application method, and nutrient release characteristics, combined with soil testing values ​​to assess the fertilizer effectiveness lag period.

[0037] The knowledge base uses a relational database to store agronomic operation threshold parameters for different tea varieties, such as the suitable pruning intensity range for Longjing 43 and the optimal nitrogen fertilizer application range for Fuding Dabai. Each data entry is associated with a growth stage marker, including key phenological periods such as budding, leaf unfolding, harvesting, and dormancy, ensuring that agronomic strategies match the tea tree's growth cycle. An operation response lag period matrix records the time it takes for different management measures to affect the physiological state of the tea trees, such as the delayed effect of root nutrient absorption after fertilization or the accumulated temperature requirement for new shoot emergence after pruning. This matrix is ​​dynamically updated through a machine learning model to improve the accuracy of predicting the effects of future agronomic operations.

[0038] The knowledge base's query and matching functions employ a similarity retrieval algorithm. When the system detects an abnormal growth status in the current tea trees, it automatically retrieves intervention records from historical data under similar environmental conditions and analyzes the actual effects of different agronomic measures. For example, if leaf edge scorching symptoms are detected, the system will retrieve treatment plans from the same season, under similar soil and weather conditions, and evaluate the improvement effects of different measures, providing a reference for current decision-making. The knowledge base's update mechanism uses incremental learning. After each agronomic intervention, the system continuously monitors changes in tea tree phenotypic changes and feeds back the actual response data to the knowledge base, optimizing the accuracy of subsequent decisions.

[0039] The implementation method of this embodiment achieves accurate analysis of the tea tree growth environment and quantitative evaluation of the effects of agronomic operations through multi-source data fusion and dynamic modeling, providing reliable data support for subsequent intelligent decision-making.

[0040] Example 2: See Figure 3 This embodiment focuses on the collaborative working mechanism of the cross-dimensional feature fusion engine and the adaptive decision optimization module. When the system initiates the consultation process, the infrared thermal imaging array first scans the tea garden area, capturing anomalous points in leaf surface temperature distribution and marking their geographic coordinates. The underground sensor group located in the anomalous area is immediately activated, uploading soil volumetric moisture content data collected over the past 72 hours at 15-minute intervals. The cumulative water stress index is calculated using the trapezoidal integral method to determine the historical water shortage level, with the formula:

[0041]

[0042] Wherein: S w θ represents the cumulative intensity of water stress experienced by tea trees over a specific time period; t θ represents the measured soil moisture content at time t. opt The optimal moisture content threshold for this variety was determined, with the integration interval [t0, t1] covering the most recent critical growth period. Visible-near-infrared spectral imagers simultaneously acquired bud and leaf images, and the distributions of brightness (L*), red-green axis (a*), and yellow-blue axis (b*) parameters were obtained through CIE-LAB color space conversion. A color matching engine retrieved the variety's baseline color gamut stored in the knowledge base and clustered and labeled pixel regions deviating from the standard value by more than 10%.

[0043] The feature cross-validation process employs a dual association mechanism. A spatiotemporal alignment protocol maps phenotypic anomaly coordinates to environmental monitoring point locations, establishing a seven-dimensional analysis model within a three-dimensional coordinate system: the X-axis represents the trajectory of soil nitrate nitrogen concentration changes, the Y-axis corresponds to the canopy's photosynthetically active radiation received, and the Z-axis represents the measured stomatal conductance. Gram angle field transform converts the environmental factor time series into feature images, and a pre-trained convolutional neural network extracts potential association patterns. When a five-day consecutive fluctuation in soil potassium content is identified as inversely related to leaf margin scorching areas, the feature association module automatically labels this area as a nutrient stress risk zone.

[0044] The construction of the multi-factor association decision tree begins with the mapping of organ function nodes in tea trees. The root vitality node is configured with five input paths: soil aeration porosity sensor data, rhizosphere pH detection sequence, mycorrhizal infection rate microscopic observation value, metabolite chromatographic analysis spectrum, and ion exchange membrane nutrient flux. The leaf photosynthesis node is associated with three dimensions of environmental parameters: calculated diffuse light ratio, ultraviolet radiation intensity distribution map, and carbon dioxide concentration gradient measurement results. The decision tree adopts a hierarchical topology. The first-level nodes contain the functions of seven key physiological organs, the second-level branches expand to 35 environmental parameter interfaces, and the third-level decision paths extend to 234 diagnostic rule chains. Each rule chain is configured with an initial weight coefficient; for example, when the available zinc content in the soil is below 2 ppm, the probability coefficient for triggering the inhibition signal at the new shoot cell division node is set to 0.83.

[0045] The adaptive decision optimization module implements dynamic weight adjustments during decision tree activation. The organ interaction analyzer continuously monitors the information transmission strength between nodes, with the initial connection weight between the root vitality node and the chlorophyll synthesis node set to 0.75. When the canopy vertical spectral index indicates increased yellowing of new leaves, the model retrieves similar phenotypic data from the historical case library and adjusts the root-leaf weight coefficient to 0.92. The tea tree growth stage adapter includes a phenological stage identification algorithm, prioritizing the terminal bud meristem node at the highest level during the budding stage and increasing the decision weight of the functional leaf node during the harvesting season. The weight iteration algorithm performs a full path evaluation every 24 hours. Specifically, if the measured daily branch elongation at a monitoring point is 15% lower than the predicted value for three consecutive days, a local reconstruction process of the decision tree is automatically triggered, deleting redundant paths with a correlation of less than 0.2 with new shoot growth and adding correlation paths between soil heat flux monitoring parameters and new shoot tissue.

[0046] The three-dimensional decision space model is established using feature mapping technology. The system projects leaf chromaticity parameters onto the X-axis coordinate interval [0,100], and the soil moisture content sequence is mapped to the Y-axis frequency domain space [0,50] after Fourier transform. The distribution of organic acid concentration in root exudates corresponds to the Z-axis scale [0,20]. When the detection points in a certain area cluster around the coordinates (62.3,17.8,8.5), spatial clustering analysis marks this area as a high-risk area for physiological water deficiency. The model uses an octree spatial segmentation algorithm, assigning a unique set of agronomic operation rules to each subspace. For example, point groups within the axial interval [50-70,10-20,5-10] are matched with emergency plans for foliar spraying of potassium humate solution.

[0047] The decision tree output mechanism employs a confidence level grading strategy. The diagnostic results for abnormal root activity are further validated with a triple verification marker: when soil oxygen diffusivity > 0.35 μg / cm³. 2 A high-confidence conclusion is output when the root tip mitotic index is >0.15 and the rate is / min; meteorological stress determination requires both leaf temperature difference >3℃ and transpiration rate decrease >40%. The time-sensitive labeling system sets execution time parameters at the end of the decision tree. For example, for a diagnosis of sudden magnesium deficiency chlorosis after rain, foliar magnesium supplementation is required at intervals not exceeding 48 hours. A decision path traceability code is automatically generated during output data encapsulation to facilitate subsequent verification of the reasoning logic chain of each diagnostic conclusion. All output results undergo multiple cross-validations before transmission: the error between short-term predicted values ​​and real-time measurement data is within 5%, and medium- and long-term trend predictions are verified through historical data backtesting. The final diagnostic report integrates the regionalized conclusions of all monitoring points, forming a zoning processing scheme including geographic grid coding.

[0048] Example 3: See Figure 4 This embodiment focuses on the generation and remote transmission mechanism of agronomic intervention plans. After receiving the diagnostic results output by the decision module, the multi-level parsing unit first processes the core parameters of the tea tree growth abnormality diagnosis results. The stress type code is converted into standard agronomic terms by a classification decoder, mapping the binary identifier to operable classification items such as "biological stress - sap-sucking pests". The severity level parameter uses a fuzzy membership function to calculate the specific degree of damage, and its numerical conversion formula is expressed as:

[0049]

[0050] Where: G is the final assessed damage level value, P represents the original detection parameters (such as insect population density or lesion coverage), α controls the slope to adjust the detection sensitivity, β sets the judgment threshold position, and γ is used for range normalization. This formula ensures that original detection values ​​of different dimensions can be uniformly converted into a standard grade score of 0-100.

[0051] The anomaly classification label generation module connects to a knowledge graph database. When the "nutrient deficiency-type chlorosis" classification label is identified, the system automatically associates the following feature combinations: interveinal chlorosis distribution pattern, signs of weakened shoot apical dominance, and color change curves associated with specific element deficiencies. The historical intervention record extraction engine activates the spatiotemporal index function to filter successful intervention cases at the same geographic grid coordinates within three years during similar phenological periods, excluding non-matching records with soil pH differences exceeding 0.5 units.

[0052] The agronomic operation combination sequence reconstructor features a dual filtering mechanism. The first-level filter retains core operation steps and removes agents from historical records that have incompatibilities with other components. The second-level filter performs timeliness checks, comparing the current phenological progress parameters of the model with the agricultural calendar within a time window constraint, and eliminating mismatched operation items. For example, it automatically blocks schemes containing growth regulators during flower bud differentiation and disables long-acting insecticides 25 days before harvest. When generating the executable operation set, three alternative schemes are retained: the basic scheme includes the lowest-cost operation unit, the optimized scheme configures synergistic adjuvants, and the emergency scheme uses fast-acting compounds.

[0053] The system optimizes the application time window by accessing a weather forecast interface to analyze data for the next 144 hours. Rainfall forecasts drive the pesticide application timing calculator, requiring at least 6 hours of no precipitation after pesticide application. A soil moisture monitoring matrix provides real-time moisture content distribution at a depth of 15cm, which the irrigation scheduler uses to calculate the soil moisture diffusion curve and determine the irrigation operation initiation threshold. When the moisture sensor indicates that the surface soil moisture content is below 60% of field capacity, the system automatically switches from liquid fertilizer application to slow-release granular application.

[0054] The structured output of the agronomic intervention program includes a five-level parameter system: core operational items define major agricultural activities such as "foliar spraying"; the material composition field records the compound name and purity; the dosage parameter is accurate to 0.01 unit concentration; the spatial execution intensity indicates the application rate per hectare; and the time sequence control marker sets the specific operation period. The dosage adjustment parameters are calculated based on the crop growth process; for example, the recommended nitrogen fertilizer amount for new shoots before harvest is dynamically adjusted according to the daily decay coefficient.

[0055] The remote consultation interface includes a device control command converter configuration protocol mapping table. The commands received by the intelligent water and fertilizer integrated machine include: solenoid valve opening time window [[10:15, 10:30]], EC value control target 1.8 mS / cm, AB solution mixing ratio 3:1, and zone execution code D7. The manual operation guidance text generation uses template filling technology, replacing the "pruning intensity" variable with a specific description of "removing 30% of inner branches".

[0056] The multimodal interaction engine's speech synthesis unit incorporates a tea cultivation terminology database, with the output frequency adjusted to the optimal audible range of 150-4500Hz. 3D motion demonstration data is achieved through a skeletal node binding algorithm, synchronizing the operator's limb movement trajectories with the virtual model at 17 key points, highlighting the correct posture where the pruning shears are angled at 45° to the main stem. The data encapsulation process employs a layered compression strategy: the basic text layer retains UTF-8 encoded instructions, the device control layer uses binary encoding, and the multimedia presentation layer utilizes the H.265 compression algorithm.

[0057] The consultation response data packet transmission protocol sets priority channels. Device control commands are pushed in real time via the MQTT protocol, with a timeliness requirement of delivery within T+0.5 hours. 3D demonstration data can be downgraded to 2D animation transmission when bandwidth is below 10Mbps. The terminal device receiving module has a command verification mechanism; when incompatibility between the device model and the command is detected, a protocol conversion service is automatically initiated, for example, converting the smart spraying vehicle coordinate commands to BeiDou grid code format. Data integrity verification uses CRC-32 cyclic redundancy check code; if verification fails, the adjacent node data reconstruction function is activated.

[0058] Example 4: This example achieves continuous optimization of the knowledge base through a closed-loop feedback mechanism. In the actual operation scenario of a tea garden, the system first deploys a composite sensor network to collect intervention execution data. Taking spring water management in the Jiangnan tea region as an example, when the system suggests implementing drip irrigation in block D7, the intelligent irrigation equipment controller records key operating parameters, as shown in Table 1.

[0059] Table 1: Sample log of agronomic intervention implementation.

[0060]

[0061]

[0062] The system synchronously activated a phenotypic change tracking mechanism. Twenty-four hours before the drip irrigation operation, the average NDVI of block D7 was 0.62 obtained by a UAV multispectral scanner, and the peak canopy temperature was recorded as 32.5℃ by a thermal imager. A follow-up examination 48 hours after the intervention showed that the NDVI had increased to 0.65, and the peak canopy temperature had decreased to 29.8℃. Leaf microscopic imaging system was used to monitor the cuticle thickness changes of the leaves in five sample groups at specific points, as shown in Table 2.

[0063] Table 2: Record of changes in phenotypic parameters before and after intervention.

[0064] Leaf thickness variation at monitoring points (μm) Daily shoot growth (mm) Pore ​​opening grade P1207+3.2 +0.36 Level 2 → Level 3 P1208+2.8 +0.41 Level 1 → Level 3 P1209+4.1 +0.39 Level 2 → Level 4 P1210+3.5 +0.32 Level 1 → Level 2 P1211+2.9 +0.35 Level 2 → Level 3

[0065] The operation response efficiency calculator initiates a multi-parameter comprehensive analysis module. Using the actual drip irrigation duration of 78 minutes, water pressure of 0.28 MPa, and EC value of 1.73 mS / cm as input variables, and the phenotypic improvement value as the output variable, a response surface model is established. The model outputs a water management efficiency coefficient of 0.87 for block D7 (1.0 is the theoretical optimal value). When this coefficient exceeds the preset threshold of 0.85, the system marks the current water management strategy as an efficient operation.

[0066] The threshold correction process is automatically activated after three consecutive efficient operations are detected. The original water stress trigger threshold in the knowledge base is 42% soil moisture content. The operation trigger points retrieved by the correction engine from the five most recent successful intervention records are shown in Table 3.

[0067] Table 3: Operation Trigger Condition Optimization Record.

[0068] Trigger Date Soil moisture content Canopy temperature ℃ New shoot growth Operation type Actual effect 20250315 43％ 33.1 0.28mm / d drip irrigation good 20250321 44％ 32.5 0.31mm / d drip irrigation excellent 20250328 45％ 32.8 0.29mm / d drip irrigation good 20250403 43％ 33.3 0.27mm / d drip irrigation excellent 20250409 44％ 32.7 0.30mm / d drip irrigation good

[0069] The revised algorithm identified that the actual effective trigger points were concentrated in the 43%-45% moisture content range. The original threshold was revised upward from 42% to 43.5%, and a canopy temperature threshold of 32.8℃ was added as a joint trigger condition. The updated parameters were immediately written into the tea tree response pattern knowledge base and marked "Verification parameters in spring 2024" on the interface.

[0070] An example of the collaborative operation of a multi-source monitoring system can be seen in the management of mountain tea gardens. A drone equipped with a hyperspectral imager performs a flight path scan every Wednesday morning, acquiring canopy reflectance data to generate an NDVI distribution map with a spatial resolution of 5 cm. A ground-based fixed sensor network comprises 36 monitoring nodes, each equipped with a six-channel spectral probe (center wavelengths: 550 nm, 650 nm, 850 nm, 1450 nm, 1650 nm, 2100 nm) to continuously record changes in leaf surface moisture content.

[0071] The rhizosphere monitoring subsystem deploys microenvironment monitoring capsules at a depth of 20cm in the roots of typical tea bushes, collecting rhizosphere exudate samples every 30 minutes. The detection data is aggregated to a regional gateway via a LoRa wireless network, forming a time-series curve of organic acid concentration in the root zone. The terrain data processor integrates satellite remote sensing imagery and field mapping data to generate a three-dimensional terrain model including slope, aspect, and elevation.

[0072] The data fusion center uses a spatiotemporal registration algorithm to align point monitoring data with areal remote sensing images. In a case study of a hilly tea garden, the system detected persistent anomalies at the monitoring point on the eastern slope (slope 22°), as shown in Table 4.

[0073] Table 4: Anomaly Analysis Records of Multi-Source Data Fusion.

[0074] Data source Monitoring parameters Normal range Measured value Deviation direction Ground node N08 650nm reflectivity 0.15-0.18 0.22 +22％ Drone aerial scanning NDVI 0.60-0.65 0.57 -5％ Rhizosphere Capsules R08 Citric acid concentration (μmol / g) 2.3-2.8 1.9 -18％ Weather station data Sunshine hours 5.2h / d 7.1h / d +37％

[0075] The 3D topology analysis engine identified the excessive sunlight problem caused by the eastern slope terrain and updated the management plan for the area: added a slope compensation factor to the moisture threshold conditions and advanced the original irrigation trigger point by 15%; created a special management strategy category for "slopes with strong sunlight" in the knowledge base; and immediately applied the optimized parameter set to areas with similar terrain.

[0076] The operational effectiveness labeling and update logic comprises a four-level verification system: Level 1 verification is based on equipment execution accuracy records; Level 2 verification relies on quantitative analysis of phenotypic changes; Level 3 verification compares the effectiveness across implementation areas; and Level 4 verification continuously tracks execution across quarters. All verified operational parameters are marked as the "golden dataset," serving as the core basis for subsequent decision-making. Invalid or inefficient parameters enter the analysis area. When a similar operation is marked as inefficient three times, a retraining process for that operation's threshold is automatically triggered.

[0077] Example 5: This example details the establishment of a multimodal tea tree phenotypic acquisition system and intelligent diagnostic triggering mechanism. The system deploys a three-dimensional sensor array at key locations in the tea garden. The depth camera is equipped with an infrared structured light projection module to record the movement trajectory of the terminal bud of the tea tree shoots with an accuracy of 0.1 mm. This device acquires spatial coordinate data every 30 seconds, constructing a three-dimensional dynamic model containing 17 biomarker points. These feature points include key locations such as the terminal bud growth cone, the attachment point of the first scale, and the base of the unfolded leaf petiole, forming a 24-hour continuous tracking curve. The morphological dynamic curve generation unit employs a polynomial fitting algorithm to quantify the changing trend of apical dominance intensity by analyzing the proportional relationship between the horizontal displacement and vertical elongation rate of the terminal bud.

[0078] The fresh leaf nutrient accumulation monitoring system integrates a high-precision weighing platform and a machine vision module. During automated harvesting, the end effector's built-in strain sensor records the weight of each cluster of fresh leaves in real time, with an accuracy of 0.01 grams. The weight data is automatically linked to timestamps and BeiDou positioning coordinates, generating a three-dimensional weight-space-time matrix. A nutrient accumulation rate calculator filters data from three consecutive harvests within the same geographic grid, generating a dynamic parameter curve based on the daily average growth formula. The system pays particular attention to the growth compensation effect of newly sprouted buds and leaves within 48 hours after harvesting, plotting a restorative growth characteristic map.

[0079] The varietal benchmark model library adopts a hierarchical storage architecture. The top-level classification is based on tea variety codes, and the secondary structure contains templates for parameters of seven key growth stages. The standard template for the spring shoot emergence period of the Longjing 43 variety includes: a daily average elongation benchmark of 1.3 mm, a leaf unfolding angle range of 35°-42°, and a second leaf area growth rate of 12.5 mm² / day. These parameters are derived from the statistical median values ​​of three consecutive years of observation, with a 5% fluctuation range threshold automatically updated at the end of each quarter. The summer shoot growth template for the Fuding Dabai variety sets different parameter combinations: peak time window for internode elongation acceleration, critical leaf age value for old leaf shedding, and lignification process coefficient of new shoots.

[0080] The anomaly detection mechanism implements a dual-channel calibration strategy. When the dynamic curve of new shoot morphology at a monitoring point deviates from the baseline model, the system activates the time series similarity assessment algorithm. Dynamic time warping technology calculates the matching degree between the actual trajectory and the template curve, setting a similarity threshold of 0.7 as the warning trigger line. For determining the material accumulation rate, a sliding window comparison algorithm is used: the average material accumulation rate of the most recent five days is selected and compared with the 25th percentile value of historical data for the same period and variety. An anomaly signal triggered in either of the dual channels activates the diagnostic request process.

[0081] The system automatically generates consultation requests with intelligent filtering conditions. Minor fluctuations caused by brief periods of insufficient light are excluded; morphological anomalies must persist for at least three consecutive monitoring cycles (4 hours per cycle). When similar anomalies occur simultaneously at multiple adjacent monitoring points, the spatial correlation analysis engine initiates cluster anomaly pattern recognition, generating only a single comprehensive request instead of multiple independent requests. The request message is encapsulated in a structured data format, with required fields including: a unique tea garden identifier (8-digit alphanumeric code), anomaly type code (e.g., G03 representing shoot growth obstruction), geographic coordinates of the center point (WGS84 coordinate system), anomaly start UTC timestamp (accurate to milliseconds), and current altitude (corrected for air pressure effects). Additional data blocks include the three most recent soil EC value monitoring records, a histogram of canopy light intensity distribution, and thumbnails of key feature point snapshots.

[0082] The request transmission protocol employs a layered encryption mechanism. The basic text layer describes the characteristics of the anomaly, such as "the vertical elongation of the terminal bud of the new shoot at monitoring point DG-07 is 1.0 mm in 24 hours, which is lower than the baseline limit of 1.2 mm." The binary data layer contains the original monitoring data packet: three-dimensional coordinates of the terminal bud at 1024 time points, 15 sets of fresh leaf weighing records, and near-infrared spectral sampling values ​​from the ambient light sensor. The metadata verification layer includes a data integrity verification code and the device calibration certificate number. All data is transmitted via 5G network slicing technology, prioritizing the bandwidth of the data stream for anomaly diagnosis requests. In signal dead zones, a local edge computing node caching mechanism is activated, and transmission automatically resumes once the network is restored.

[0083] The system maintenance module implements a periodic self-calibration process. Sensor drift detection is initiated every hour on the hour, and a standard calibration board is moved to the center of the camera's field of view by a robotic arm for baseline value verification. When a positioning deviation of the infrared structured light projection point exceeds 50 micrometers, a compensation algorithm is automatically invoked to update the intrinsic parameter matrix. The weighing system calibration uses an automatic standard weight loading device; if the error exceeds 0.05 grams, a level three alarm is triggered, and data acquisition is suspended. All calibration records are synchronously written to the metadata layer of the consultation request to ensure accurate assessment of data reliability during diagnostic analysis.

[0084] The data preprocessing workflow includes intelligent outlier filtering. When data from a single monitoring point suddenly jumps by more than three standard deviations, the system automatically checks for environmental interference sources: whether there have been infrared sensor records of people approaching within the past five minutes, whether the weather station has reported a sudden strong wind event, and whether adjacent sensors have experienced synchronous disturbances. Interfering data points confirmed not to be related to plant physiological changes are marked as invalid values, and automatic interpolation compensation is performed during the preprocessing stage. Only when coordinated changes occur at multiple consecutive monitoring points is the system considered a valid physiological response signal and proceeds to the subsequent analysis process. Each transmitted consultation request is accompanied by a data quality control report, detailing the original acquisition quality and correction history for each parameter.

[0085] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0086] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A remote consultation method for tea experts based on artificial intelligence, characterized in that, Includes the following steps: Acquire tea tree growth environment data and tea garden management operation records. The tea tree growth environment data includes soil physicochemical indicators, temporal changes of meteorological factors, and biological stress signals. The tea tree physiological state dynamic analysis module processes the tea tree growth environment data to generate a tea tree physiological state feature vector. Based on the tea garden management operation records, historical agronomic operation nodes are extracted and associated with the tea tree physiological state feature vector to construct a tea tree response pattern knowledge base. Consultation requests are triggered based on real-time collected tea tree phenotypic data. A cross-dimensional feature fusion engine integrates the tea tree physiological state feature vector with the tea tree response pattern knowledge base to generate a multi-factor association decision tree. An adaptive decision optimization module iteratively adjusts the node weights of the multi-factor association decision tree, outputting tea tree growth anomaly diagnosis results and agronomic intervention plans. The agronomic intervention plans are transmitted to the terminal device via a remote consultation interface, while simultaneously updating the operation validity markers in the tea tree response pattern knowledge base.

2. The remote tea expert consultation method based on artificial intelligence according to claim 1, characterized in that, The process of processing tea tree growth environment data through a dynamic analysis module of tea tree physiological state to generate a tea tree physiological state feature vector includes: separating time-series data of nitrogen, phosphorus, and potassium content and spatial distribution data of trace elements from the soil physicochemical indicators, and inputting them into a soil nutrient response model to generate soil fertility characteristics; analyzing the light intensity fluctuation pattern and temperature and humidity co-change curve in the time-series changes of meteorological factors, and outputting a meteorological stress index through a microclimate impact assessment model; identifying the spectrum of pest and disease characteristics and the activity trajectory of natural enemies in the biological stress signals, and generating ecological balance parameters using a biological interaction map; and fusing the soil fertility characteristics, meteorological stress index, and ecological balance parameters, and generating a tea tree physiological state feature vector through multi-dimensional feature dimensionality reduction processing.

3. The method for remote consultation with tea experts based on artificial intelligence according to claim 2, characterized in that, The construction of a tea tree response pattern knowledge base by associating the tea tree physiological state feature vector includes: matching pruning time markers and fertilizer application change records in the historical agronomic operation nodes to extract the time series sequence of agronomic operation intensity; establishing the dynamic response relationship between the time series sequence of agronomic operation intensity and the tea tree physiological state feature vector to generate an operation response lag period matrix; processing the operation response lag period matrix through a tea tree variety adaptability analysis model to classify and store agronomic operation threshold parameters for different tea tree varieties; and integrating the operation response lag period matrix and the agronomic operation threshold parameters to construct a tea tree response pattern knowledge base containing tea tree growth stage markers.

4. The method for remote consultation with tea experts based on artificial intelligence according to claim 3, characterized in that, The process of integrating the tea tree physiological state feature vector with the tea tree response pattern knowledge base through a cross-dimensional feature fusion engine to generate a multi-factor association decision tree includes: extracting bud and leaf color distribution parameters and branch elongation rate from the current tea tree phenotypic data to generate phenotypic anomaly feature codes; spatiotemporally aligning the phenotypic anomaly feature codes with the tea tree physiological state feature vector, and generating environmental stress association factors through a feature cross-validation module; matching the growth stage markers of the same tea tree variety in the tea tree response pattern knowledge base, and calling the corresponding agronomic operation threshold parameters; and fusing the environmental stress association factors and the agronomic operation threshold parameters to construct a multi-factor association decision tree with tea tree organs as nodes.

5. The method for remote consultation with tea experts based on artificial intelligence according to claim 4, characterized in that, The step of using an adaptive decision optimization module to iteratively adjust the node weights of the multi-factor association decision tree includes: identifying the strength of the association path between leaf nodes and root nodes in the multi-factor association decision tree and generating organ interaction weight coefficients; correcting the initial values ​​of the organ interaction weight coefficients based on intervention effect feedback data in a historical consultation case library; adjusting the decision priority of different organ nodes through an adaptive function for tea tree growth stages and generating a dynamic weight allocation table; and reorganizing the nodes of the multi-factor association decision tree according to the dynamic weight allocation table to output a diagnostic result of tea tree growth abnormality containing time-sensitive markers.

6. The method for remote consultation with tea experts based on artificial intelligence according to claim 5, characterized in that, The output of abnormal tea tree growth diagnosis results and agronomic intervention plans includes: parsing the stress type code and severity level parameters in the abnormal tea tree growth diagnosis results to generate abnormal classification labels; associating historical intervention records with the same abnormal classification labels in the tea tree response pattern knowledge base to extract agronomic operation combination sequences; processing the agronomic operation combination sequences through a timeliness constraint model to select executable operation sets that conform to the current phenological period; integrating weather forecast data and real-time soil moisture monitoring values ​​to optimize the implementation time window of the executable operation sets and generate an agronomic intervention plan that includes dosage adjustment parameters.

7. The method for remote consultation with tea experts based on artificial intelligence according to claim 6, characterized in that, The step of transmitting the agronomic intervention plan to the terminal device via a remote consultation interface includes: decomposing the operation steps in the agronomic intervention plan into equipment control instructions and manual operation guidance text; adapting the equipment control instructions to the communication protocol of the intelligent irrigation equipment and fertilizer machinery through a protocol conversion gateway; synthesizing voice prompts and three-dimensional motion demonstration data of the manual operation guidance text using a multimodal interaction engine; and encapsulating the equipment control instructions, voice prompts, and three-dimensional motion demonstration data to form a consultation response data packet for transmission to the terminal device.

8. The method for remote consultation with tea experts based on artificial intelligence according to claim 7, characterized in that, The step of updating the operation validity marker of the tea tree response pattern knowledge base includes: collecting agronomic intervention execution logs fed back by terminal devices and analyzing actual operation time deviation and dosage deviation data; comparing the change rate of the tea tree phenotypic data before and after intervention to calculate the operation response efficiency coefficient; associating the original agronomic operation threshold parameters in the tea tree response pattern knowledge base with the operation trigger condition threshold and correcting the operation trigger condition threshold; binding the operation response efficiency coefficient with the corrected operation trigger condition threshold and updating the operation validity marker of the tea tree response pattern knowledge base.

9. The method for remote consultation with tea experts based on artificial intelligence according to claim 1, characterized in that, Before acquiring tea tree growth environment data and tea garden management operation records, the process also includes: deploying multispectral sensing devices to collect tea tree canopy reflectance data and generating a spatial distribution map of chlorophyll content; installing underground rhizosphere monitoring probes to acquire time-series data of root exudate concentration and constructing root zone microbial activity indicators; and using a wireless sensor network to aggregate the spatial distribution map of chlorophyll content and the root zone microbial activity indicators, and fusing topographic relief parameters from satellite remote sensing images to generate a basic dataset of tea tree growth environment.

10. The method for remote consultation with tea experts based on artificial intelligence according to claim 9, characterized in that, The process of triggering consultation requests based on real-time collected tea tree phenotypic data includes: capturing the sequence of changes in the growth angle of new tea tree shoots using a depth camera device to generate a morphogenesis dynamic curve; measuring the weight increment of fresh leaves using a high-precision weighing sensor and generating a substance accumulation rate parameter in combination with the harvesting timestamp; and automatically generating a consultation request containing a geolocation code when the morphogenesis dynamic curve deviates from the variety benchmark model or the substance accumulation rate parameter is lower than a threshold.

Images