Tea garden drought visual monitoring method and system

Abstract

The invention provides a visual monitoring method and system for tea garden drought in the technical field of intelligent agriculture. The method comprises the steps that S1, a large amount of historical monitoring data is collected from a tea garden through a sensor array; s2, preprocessing each historical monitoring data and then constructing a data set; s3, analyzing the data set through an L-imma algorithm, a COX survival regression model and a GLM generalized linear model to obtain an environmental factor correlation graph; s4, creating a tea garden drought prediction model based on the environmental factor correlation graph; s5, training and deploying the tea garden drought prediction model through the data set; s6, collecting real-time monitoring data from the tea garden, and inputting the data into the tea garden drought prediction model to obtain a tea garden drought prediction result; and S7, converting the tea garden drought prediction result into a drought thermodynamic diagram for display. The method has the advantages that the accuracy of tea garden drought prediction, the refinement degree of irrigation decision and the intuition of drought display are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of smart agricultural technology, and in particular to a method and system for visually monitoring drought conditions in tea gardens. Background Art

[0002] Drought has a significant impact on tea plant growth, physiological metabolism, tea quality, and yield. Drought stress can have at least the following effects on tea plants: 1. It causes tea leaves to become smaller and thinner, with shortened leaf length and width, fewer leaves, and dwarfed plants. 2. It increases the number of lateral roots, but restricts overall root development, causing root hairs to dry out and die, and reducing absorption capacity. 3. Long-term drought can lead to stagnation of tea plant growth, slow new shoot growth, bud and leaf shrinkage, and even death. 4. It inhibits photosynthesis, resulting in a reduced photosynthetic rate. 5. It can cause changes in the content of caffeine, amino acids, tea polyphenols, and other quality components in tea leaves. Therefore, it is necessary to predict drought conditions in tea plantations for better management. Soil moisture, which refers to the content, distribution, and variation of soil water, is a key environmental indicator reflecting soil drought conditions and directly affects tea plant growth. Predicting soil moisture can help predict drought conditions and allow irrigation to be implemented at the appropriate time, saving water resources and ensuring healthy tea plant growth.

[0003] Traditionally, the prediction of drought in tea gardens has simply involved collecting some tea garden monitoring data through sensors, then using the data set constructed from the tea garden monitoring data to train a neural network model, and using the trained neural network model to determine whether there is a drought. However, this method has the following shortcomings: 1. Since the factors with the highest correlation with drought in the tea garden monitoring data are not screened, the model will be affected by factors with low correlation with drought during the prediction process, resulting in poor prediction accuracy; 2. It only simply predicts whether there is a drought, and does not implement hierarchical management of the drought, resulting in the inability to carry out refined irrigation in the tea garden and a waste of water resources; 3. The information presentation method is single, and is only based on simple push notifications via text messages. Managers cannot accurately determine information such as the spatial distribution of the drought through text messages.

[0004] Therefore, how to provide a visual monitoring method and system for tea garden drought to improve the accuracy of tea garden drought prediction, the refinement of irrigation decision-making, and the intuitiveness of drought display has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for visually monitoring drought conditions in tea gardens, so as to improve the accuracy of drought prediction in tea gardens, the refinement of irrigation decision-making, and the intuitiveness of drought display.

[0006] In a first aspect, the present invention provides a method for visually monitoring drought conditions in a tea garden, comprising the following steps:

[0007] Step S1: collecting a large amount of historical monitoring data from the tea garden through a sensor array;

[0008] Step S2: constructing a data set after preprocessing each of the historical monitoring data, and performing a sample expansion operation on the data set;

[0009] Step S3, analyzing the data set using the Limma algorithm, the COX survival regression model, and the GLM generalized linear model to obtain an environmental factor correlation graph;

[0010] Step S4: creating a tea garden drought prediction model based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer, and drought mapping output layer, and setting a loss function of the tea garden drought prediction model;

[0011] The multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features, and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features according to the environmental factor correlation diagram, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction result with the drought period, drought level, and drought location;

[0012] Step S5, dividing the data set into a training set, a validation set, and a test set, training the tea garden drought prediction model using the training set and a loss function, validating the trained tea garden drought prediction model using the validation set, testing the verified tea garden drought prediction model using the test set, and deploying the tested tea garden drought prediction model using hardware acceleration technology;

[0013] Step S6: collecting real-time monitoring data from the tea garden through the sensor array, pre-processing the real-time monitoring data, and inputting the data into the deployed tea garden drought prediction model to obtain a tea garden drought prediction result;

[0014] Step S7: converting the tea garden drought prediction result into a drought heat map, and displaying the drought heat map in real time through a visual interface.

[0015] Furthermore, the step S1 is specifically as follows:

[0016] A large amount of historical monitoring data is collected from the tea garden through a sensor array including humidity sensors, light intensity sensors, temperature sensors, carbon dioxide sensors, air pressure sensors, particulate matter sensors, global solar radiation sensors, pH sensors, salinity sensors, conductivity sensors, rainfall sensors, wind speed and direction sensors, and positioners;

[0017] The historical monitoring data at least include maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, daily temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure value, minimum air pressure value, average air pressure value, average daytime air pressure value, average nighttime air pressure value, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ total radiation, minimum TBQ total radiation, daily average TBQ total radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity value, minimum soil salinity value, average soil salinity value, maximum soil conductivity, minimum soil conductivity, average soil conductivity, rainfall, wind direction, wind speed, sampling location and sampling time.

[0018] Furthermore, the step S2 is specifically as follows:

[0019] After filling missing values ​​and repairing outliers for each of the historical monitoring data, each of the historical monitoring data is labeled with soil entropy, drought level, and time segment labels to complete the preprocessing of each of the historical monitoring data. A data set is constructed based on the preprocessed historical monitoring data, and a sample expansion operation is performed on the data set through an autoregressive model; the time segment label is the month of the sampling time;

[0020] The missing value filling is based on the K-nearest neighbor filling method; the outlier repair is based on the Z-score algorithm and the K-means clustering algorithm, specifically:

[0021] Perform Z-score standardization on each of the historical monitoring data, calculate the Z-score value of each of the historical monitoring data after Z-score standardization, and compare the Z-score value with a preset abnormality threshold to screen out abnormal values; cluster the historical monitoring data after Z-score standardization using a K-means clustering algorithm to obtain several cluster centers, and repair each abnormal value in turn based on the nearest cluster center.

[0022] Furthermore, the step S3 is specifically as follows:

[0023] Segmenting the data set based on time segmentation labels to obtain a number of data subsets, and screening significant factors between the data subsets whose change rates exceed a preset change threshold using the Limma algorithm;

[0024] Calculating a first drought risk level for each of the significant factors using a COX survival regression model, calculating a second drought risk level for each of the significant factors using a GLM generalized linear model, cross-validating the first drought risk level and the second drought risk level to screen N environmental factors with the greatest impact on drought conditions under different time segment labels from each of the significant factors, where N is a positive integer; and constructing an environmental factor correlation graph based on the screened environmental factors and time segment labels.

[0025] In step S4, the multimodal feature extraction layer is constructed based on the meteorological time series encoding module, the soil dynamic encoding module and the spatiotemporal embedding module;

[0026] The meteorological time series encoding module is used to extract time-dependent features from the monitoring data through a bidirectional LSTM network; the soil dynamic encoding module is used to extract soil dynamic features from the monitoring data through a fully connected network with residual connections; the spatiotemporal embedding module is used to extract spatiotemporal features from the monitoring data through an embedding layer;

[0027] The feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features by combining the environmental factor correlation graph with a multi-head attention mechanism, and output multimodal features through 1D convolutional layer fusion;

[0028] The soil entropy prediction layer is used to infer multimodal features through gated recurrent units and feature decoupling heads to calculate soil entropy values;

[0029] The specific drought classification rules are as follows:

[0030] The quantile of soil entropy value is (75,100], which corresponds to the first level of drought; the quantile is (25,75], which corresponds to the second level of drought; the quantile is (5,25], which corresponds to the third level of drought; the quantile is (0,5], which corresponds to the fourth level of drought.

[0031] Furthermore, the step S5 is specifically as follows:

[0032] Using a stratified sampling technique, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The tea garden drought prediction model is trained using the training set. During the training process, the hyperparameters of the tea garden drought prediction model, including at least a learning rate, a batch size, and a number of training rounds, are continuously optimized until the loss function converges or a preset convergence condition is reached;

[0033] The trained tea garden drought prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the verification fails, and the training set is expanded to continue training. If so, the verification passes, and:

[0034] Calculating the recall rate, F1 value and mean square error of the test set to test the tea garden drought prediction model, and deploying the tea garden drought prediction model that passes the test through hardware acceleration technology;

[0035] The step S7 is specifically as follows:

[0036] The heat map generation engine converts the tea garden drought prediction results into a drought heat map through an adaptive color scale mapping algorithm and a multi-scale visualization strategy, loads the drought heat map onto a GIS map, and displays it in real time through a visualization interface. The visualization interface refreshes the drought heat map based on a preset refresh cycle.

[0037] In a second aspect, the present invention provides a tea garden drought visualization monitoring system, comprising the following modules:

[0038] A historical monitoring data collection module is used to collect a large amount of historical monitoring data from the tea garden through a sensor array;

[0039] A data set expansion module, configured to construct a data set after preprocessing each of the historical monitoring data, and perform a sample expansion operation on the data set;

[0040] An environmental factor correlation graph generation module is used to analyze the data set using the Limma algorithm, the COX survival regression model, and the GLM generalized linear model to obtain an environmental factor correlation graph;

[0041] A tea garden drought prediction model creation module is used to create a tea garden drought prediction model based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer, and drought mapping output layer, and set the loss function of the tea garden drought prediction model;

[0042] The multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features, and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features according to the environmental factor correlation diagram, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction result with the drought period, drought level, and drought location;

[0043] A tea garden drought prediction model training module is used to divide the data set into a training set, a validation set, and a test set, train the tea garden drought prediction model using the training set and a loss function, verify the trained tea garden drought prediction model using the validation set, test the verified tea garden drought prediction model using the test set, and deploy the tested tea garden drought prediction model using hardware acceleration technology;

[0044] A tea garden drought prediction module is used to collect real-time monitoring data from the tea garden through a sensor array, pre-process the real-time monitoring data, and input the data into the deployed tea garden drought prediction model to obtain a tea garden drought prediction result;

[0045] The tea garden drought visualization display module is used to convert the tea garden drought prediction results into a drought heat map and display the drought heat map in real time through a visualization interface.

[0046] Furthermore, the historical monitoring data collection module is specifically used to:

[0047] A large amount of historical monitoring data is collected from the tea garden through a sensor array including humidity sensors, light intensity sensors, temperature sensors, carbon dioxide sensors, air pressure sensors, particulate matter sensors, global solar radiation sensors, pH sensors, salinity sensors, conductivity sensors, rainfall sensors, wind speed and direction sensors, and positioners;

[0048] The historical monitoring data at least include maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, daily temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure value, minimum air pressure value, average air pressure value, average daytime air pressure value, average nighttime air pressure value, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ total radiation, minimum TBQ total radiation, daily average TBQ total radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity value, minimum soil salinity value, average soil salinity value, maximum soil conductivity, minimum soil conductivity, average soil conductivity, rainfall, wind direction, wind speed, sampling location and sampling time.

[0049] Furthermore, the data set expansion module is specifically used to:

[0050] After filling missing values ​​and repairing outliers for each of the historical monitoring data, each of the historical monitoring data is labeled with soil entropy, drought level, and time segment labels to complete the preprocessing of each of the historical monitoring data. A data set is constructed based on the preprocessed historical monitoring data, and a sample expansion operation is performed on the data set through an autoregressive model; the time segment label is the month of the sampling time;

[0051] The missing value filling is based on the K-nearest neighbor filling method; the outlier repair is based on the Z-score algorithm and the K-means clustering algorithm, specifically:

[0052] Perform Z-score standardization on each of the historical monitoring data, calculate the Z-score value of each of the historical monitoring data after Z-score standardization, and compare the Z-score value with a preset abnormality threshold to screen out abnormal values; cluster the historical monitoring data after Z-score standardization using a K-means clustering algorithm to obtain several cluster centers, and repair each abnormal value in turn based on the nearest cluster center.

[0053] Furthermore, the environmental factor correlation graph generation module is specifically used to:

[0054] Segmenting the data set based on time segmentation labels to obtain a number of data subsets, and screening significant factors between the data subsets whose change rates exceed a preset change threshold using the Limma algorithm;

[0055] Calculating a first drought risk level for each of the significant factors using a COX survival regression model, calculating a second drought risk level for each of the significant factors using a GLM generalized linear model, cross-validating the first drought risk level and the second drought risk level to screen N environmental factors with the greatest impact on drought conditions under different time segment labels from each of the significant factors, where N is a positive integer; and constructing an environmental factor correlation graph based on the screened environmental factors and time segment labels.

[0056] In the tea garden drought prediction model creation module, the multimodal feature extraction layer is constructed based on the meteorological time series coding module, the soil dynamic coding module and the spatiotemporal embedding module;

[0057] The meteorological time series encoding module is used to extract time-dependent features from the monitoring data through a bidirectional LSTM network; the soil dynamic encoding module is used to extract soil dynamic features from the monitoring data through a fully connected network with residual connections; the spatiotemporal embedding module is used to extract spatiotemporal features from the monitoring data through an embedding layer;

[0058] The feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features by combining the environmental factor correlation graph with a multi-head attention mechanism, and output multimodal features through 1D convolutional layer fusion;

[0059] The soil entropy prediction layer is used to infer multimodal features through gated recurrent units and feature decoupling heads to calculate soil entropy values;

[0060] The specific drought classification rules are as follows:

[0061] The quantile of soil entropy value is (75,100], which corresponds to the first level of drought; the quantile is (25,75], which corresponds to the second level of drought; the quantile is (5,25], which corresponds to the third level of drought; the quantile is (0,5], which corresponds to the fourth level of drought.

[0062] Furthermore, the tea garden drought prediction model training module is specifically used to:

[0063] Using a stratified sampling technique, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The tea garden drought prediction model is trained using the training set. During the training process, the hyperparameters of the tea garden drought prediction model, including at least a learning rate, a batch size, and a number of training rounds, are continuously optimized until the loss function converges or a preset convergence condition is reached;

[0064] The trained tea garden drought prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the verification fails, and the training set is expanded to continue training. If so, the verification passes, and:

[0065] Calculating the recall rate, F1 value and mean square error of the test set to test the tea garden drought prediction model, and deploying the tea garden drought prediction model that passes the test through hardware acceleration technology;

[0066] The tea garden drought visualization display module is specifically used for:

[0067] The heat map generation engine converts the tea garden drought prediction results into a drought heat map through an adaptive color scale mapping algorithm and a multi-scale visualization strategy, loads the drought heat map onto a GIS map, and displays it in real time through a visualization interface. The visualization interface refreshes the drought heat map based on a preset refresh cycle.

[0068] The advantages of the present invention are:

[0069] 1. A large amount of historical monitoring data is collected from the tea garden through the sensor array to construct a data set, and a sample expansion operation is performed on the data set; then the data set is analyzed by the Limma algorithm, COX survival regression model and GLM generalized linear model to obtain an environmental factor correlation map; then a tea garden drought prediction model is created based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer and drought mapping output layer, and the loss function of the tea garden drought prediction model is set; the multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features and spatiotemporal features according to the environmental factor correlation map, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction results with drought period, drought level and drought location; then the data set is divided into training set, validation set and test set. A test set is used to train, verify, and test the tea garden drought prediction model. The tea garden drought prediction model that passes the test is deployed using hardware acceleration technology. Then, real-time monitoring data is collected from the tea garden through a sensor array. After preprocessing, the real-time monitoring data is input into the deployed tea garden drought prediction model to obtain the tea garden drought prediction results. The tea garden drought prediction results are converted into drought heat maps for real-time display through a visualization interface. That is, the tea garden drought prediction model is used to predict the tea garden drought. Before the prediction, the feature fusion layer of the tea garden drought prediction model calls the environmental factor correlation map to adjust the weight of each feature. That is, the weight of the features corresponding to the N environmental factors with the greatest impact on drought prediction under the current time segment label is increased. The tea garden drought prediction results are output in combination with the drought classification rules to facilitate refined irrigation of the tea garden. The tea garden drought prediction results are converted into drought heat maps for visualization, which can accurately judge the spatial distribution of drought, ultimately greatly improving the accuracy of tea garden drought prediction, the refinement of irrigation decision-making, and the intuitiveness of drought display.

[0070] 2. By using an array of 13 types of sensors (such as humidity, light, CO2, etc.) to collect data, covering multi-dimensional parameters such as meteorology, soil, and environmental pollutants (such as maximum / average / daytime and nighttime sub-item data), dynamic monitoring of all elements of the tea garden microenvironment is achieved, significantly improving data comprehensiveness and monitoring accuracy.

[0071] 3. By combining Z-score anomaly detection and K-means clustering repair: First, identify outliers by quantifying the degree of deviation through Z-score, and then use the cluster center value to repair them. Compared with traditional single methods, this method can better maintain the data distribution characteristics and avoid sample bias caused by simple elimination.

[0072] 4. By using the autoregressive model for sample expansion, the statistical characteristics of the time series are maintained, the problem of insufficient seasonal samples in agricultural data is solved, and the generalization ability of the model is improved.

[0073] 5. After screening significant factors using the Limma algorithm, the risk level is calculated using both COX survival regression (time-related risk model) and GLM generalized linear model (nonlinear relationship modeling). Cross-validation is used to ensure the statistical significance of environmental factor screening and effectively reduce the risk of misselection of a single model.

[0074] 6. Constructing an environmental factor correlation diagram by time segmentation (such as month) can capture the time-varying characteristics of the relationship between environmental factors and drought conditions. Compared with static correlation analysis, it is more in line with the seasonal laws of agricultural production.

[0075] 7. Meteorological time series coding (bidirectional LSTM) is used to capture the long-term dependence of meteorological parameters such as light and temperature. Soil dynamic coding (residual FCN) is used to learn the mutation characteristics of indicators such as soil pH and salinity. The spatiotemporal embedding module fuses the joint features of geographic location and timestamp. In other words, the parallel extraction of three types of heterogeneous features fully explores the potential correlation of the data, thereby greatly improving the accuracy of tea garden drought prediction.

[0076] 8. Through the multi-head attention mechanism, combined with the environmental factor correlation graph, the feature weights are dynamically adjusted, so that the model can automatically focus on key features according to different time periods and regional characteristics, thereby improving the effectiveness of feature combinations.

[0077] 9. By setting the soil moisture prediction layer to use GRU+feature decoupling head, it not only retains the gated transmission of sequence information, but also separates redundant noise through decoupling operations, which is more interpretable than the traditional end-to-end prediction structure.

[0078] 10. Maintaining data distribution consistency through stratified sampling at a ratio of 8:1:1, improving training efficiency through automatic optimization of hyperparameters (learning rate, batch size, etc.), and ensuring real-time prediction through hardware acceleration deployment to meet the timeliness requirements of tea garden monitoring.

[0079] 11. By adopting an adaptive color scale mapping algorithm to dynamically adjust the scale range of the thermal map, combining multi-scale visualization strategies (such as regional / plot / plant level views), and cooperating with GIS maps to achieve precise spatial positioning, it supports multi-granularity drought analysis from macro trends to micro areas.

[0080] 12. By forming a complete technical closed loop from data collection → preprocessing → model training → visualization, the synergistic effect of innovations in each link makes the overall solution significantly superior to traditional monitoring methods in key indicators such as prediction accuracy (F1 value, mean square error), result interpretability, and system response speed.

[0081] 13. Locators collect sampling location data and dynamically bind it to GIS maps, enabling precise spatial mapping of drought forecast results. Compared to traditional methods that simply mark plot numbers, this technology can accurately map latitude and longitude coordinates, providing centimeter-level positioning support for precision agriculture operations such as drone irrigation.

[0082] 14. By using months as time segment labels (rather than simple quarterly or annual divisions), it is more in line with the growth cycle of tea trees (such as the spring tea picking period, summer dormancy period, etc.), so that the results of the environmental factor correlation analysis are highly matched with the physiological stages of tea trees, thereby improving the agronomic rationality of drought prediction.

[0083] 15. By innovatively applying the Limma algorithm for differential gene screening in bioinformatics to the significance analysis of environmental factors, and taking advantage of its linear model fitting and Bayesian statistical advantages, we can efficiently identify sensitive parameters that change over time in the tea garden microenvironment (such as day and night temperature differences and CO2 concentration fluctuations), breaking through the limitations of traditional variance analysis.

[0084] 16. By setting the soil dynamic encoding module to adopt a residual fully connected network (instead of a complex CNN), the number of parameters can be reduced while avoiding gradient vanishing; combined with the GRU gated recurrent unit (lighter than LSTM), a balance between high-precision prediction and low computing resources is achieved, and it is suitable for edge computing devices (such as field embedded terminals).

[0085] 17. Dynamic monitoring of all elements of the tea garden environment is achieved through a multi-dimensional sensor array, and data quality is improved by combining dual anomaly repair with Z-score and K-means and autoregressive data expansion; the Limma algorithm, COX survival regression and GLM model are innovatively integrated to screen time-varying environmental factors, and a multimodal deep learning model (bidirectional LSTM, residual network, spatiotemporal embedding) is constructed to extract heterogeneous features. The features are dynamically fused based on the adaptive attention mechanism, supplemented by quantile dynamic classification rules and hardware acceleration deployment, to achieve high-precision soil moisture prediction; at the same time, through the multi-scale heat map visualization and closed-loop iteration mechanism integrated by GIS, taking into account both macro trend analysis and micro precise positioning, a full-chain technological innovation is formed from data collection, model reasoning to decision support, which significantly improves the timeliness, accuracy and adaptability of drought monitoring to agricultural scenarios, and has the dual advantages of interdisciplinary method integration and engineering implementation. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0087] Figure 1 The present invention is a flow chart of a method for visually monitoring drought conditions in tea gardens.

[0088] Figure 2 The present invention is a structural diagram of a visual monitoring system for drought conditions in tea gardens. DETAILED DESCRIPTION

[0089] The technical solution in the embodiment of the present application has the following overall idea: the tea garden drought is predicted through a pre-trained tea garden drought prediction model. Before the prediction, the environmental factor correlation map is called through the feature fusion layer of the tea garden drought prediction model to adjust the weight of each feature, that is, the weights of the features corresponding to the N environmental factors that have the greatest impact on the drought prediction under the current time segment label are increased, and the tea garden drought prediction results are output in combination with the drought classification rules to facilitate the refined irrigation of the tea garden. The tea garden drought prediction results are converted into a drought heat map for visual display, which can accurately judge the spatial distribution of the drought, thereby improving the accuracy of the tea garden drought prediction, the refinement of the irrigation decision, and the intuitiveness of the drought display.

[0090] Please refer to Figures 1 to 2 As shown, a preferred embodiment of a method for visually monitoring drought conditions in a tea garden according to the present invention comprises the following steps:

[0091] Step S1: collecting a large amount of historical monitoring data from the tea garden through a sensor array;

[0092] Step S2: constructing a data set after preprocessing each of the historical monitoring data, and performing a sample expansion operation on the data set;

[0093] Step S3, analyzing the data set using the Limma algorithm, the COX survival regression model, and the GLM generalized linear model to obtain an environmental factor correlation graph;

[0094] Step S4: creating a tea garden drought prediction model based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer, and drought mapping output layer, and setting a loss function of the tea garden drought prediction model;

[0095] The multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features and spatiotemporal features according to the environmental factor correlation diagram, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction result with the drought period, drought level and drought location; the drought mapping output layer outputs the probability of the drought level (level 1 drought / level 2 drought / level 3 drought / level 4 drought), the probability of the drought period, and the probability of the drought location through the fully connected layer + Softmax, and introduces the temperature scaling technology to calibrate the prediction confidence;

[0096] Step S5: dividing the data set into a training set, a validation set, and a test set; training the tea garden drought prediction model using the training set and a loss function; validating the trained tea garden drought prediction model using the validation set; testing the verified tea garden drought prediction model using the test set; and deploying the tested tea garden drought prediction model using hardware acceleration technology; using the AdamW optimizer in the training process, combined with cosine annealing learning rate scheduling;

[0097] Step S6: collecting real-time monitoring data from the tea garden through the sensor array, pre-processing the real-time monitoring data, and inputting the data into the deployed tea garden drought prediction model to obtain a tea garden drought prediction result;

[0098] Step S7: converting the tea garden drought prediction result into a drought heat map, and displaying the drought heat map in real time through a visual interface.

[0099] By forming a complete technical closed loop from data collection → preprocessing → model training → visualization, the synergistic effect of innovations in each link makes the overall solution significantly superior to traditional monitoring methods in key indicators such as prediction accuracy (F1 value, mean square error), result interpretability, and system response speed.

[0100] Dynamic monitoring of all elements of the tea garden environment is achieved through a multi-dimensional sensor array, and data quality is improved by combining dual anomaly repair with Z-score and K-means and autoregressive data expansion. The Limma algorithm, COX survival regression and GLM model are innovatively integrated to screen time-varying environmental factors, and a multimodal deep learning model (bidirectional LSTM, residual network, spatiotemporal embedding) is constructed to extract heterogeneous features. The features are dynamically fused based on an adaptive attention mechanism, supplemented by quantile dynamic classification rules and hardware acceleration deployment, to achieve high-precision soil moisture prediction. At the same time, through GIS integrated multi-scale heat map visualization and closed-loop iteration mechanism, macro trend analysis and micro precise positioning are taken into account, forming a full-chain technical innovation from data collection, model reasoning to decision support, significantly improving the timeliness, accuracy and adaptability of drought monitoring to agricultural scenarios, and having the dual advantages of interdisciplinary method integration and engineering implementation.

[0101] The step S1 is specifically as follows:

[0102] A large amount of historical monitoring data is collected from the tea garden through a sensor array including humidity sensors, light intensity sensors, temperature sensors, carbon dioxide sensors, air pressure sensors, particulate matter sensors, global solar radiation sensors, pH sensors, salinity sensors, conductivity sensors, rainfall sensors, wind speed and direction sensors, and positioners;

[0103] By using an array of 13 types of sensors (such as humidity, light, CO2, etc.) to collect data, covering multi-dimensional parameters such as meteorology, soil, and environmental pollutants (such as maximum / average / daytime and nighttime sub-item data), dynamic monitoring of all elements of the tea garden microenvironment is achieved, significantly improving data comprehensiveness and monitoring accuracy.

[0104] By collecting sampling location data through locators and dynamically binding it to GIS maps, drought forecasts can be accurately mapped to spatial coordinates. Compared to traditional methods that only mark plot numbers, this technology can accurately determine longitude and latitude coordinates, providing centimeter-level positioning support for precision agriculture operations such as drone irrigation.

[0105] The historical monitoring data at least include maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, daily temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure value, minimum air pressure value, average air pressure value, average daytime air pressure value, average nighttime air pressure value, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ total radiation, minimum TBQ total radiation, daily average TBQ total radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity value, minimum soil salinity value, average soil salinity value, maximum soil conductivity, minimum soil conductivity, average soil conductivity, rainfall, wind direction, wind speed, sampling location and sampling time.

[0106] The step S2 is specifically as follows:

[0107] After filling missing values ​​and repairing outliers for each of the historical monitoring data, each of the historical monitoring data is labeled with soil entropy, drought level, and time segment labels to complete the preprocessing of each of the historical monitoring data. A data set is constructed based on the preprocessed historical monitoring data, and a sample expansion operation is performed on the data set through an autoregressive model; the time segment label is the month of the sampling time;

[0108] By using the autoregressive model to expand the sample, we can not only maintain the statistical characteristics of the time series, but also solve the problem of insufficient seasonal samples in agricultural data and improve the generalization ability of the model.

[0109] By using months as time segment labels (rather than simple quarterly or annual divisions), it is more in line with the growth cycle of tea trees (such as the spring tea picking period, summer dormancy period, etc.), so that the results of the environmental factor correlation analysis are highly matched with the physiological stages of tea trees, thereby improving the agronomic rationality of drought prediction.

[0110] The missing value filling is based on the K-nearest neighbor filling method; the outlier repair is based on the Z-score algorithm and the K-means clustering algorithm, specifically:

[0111] Perform Z-score standardization on each of the historical monitoring data, calculate the Z-score value of each of the historical monitoring data after Z-score standardization, and compare the Z-score value with a preset abnormality threshold to screen out abnormal values; cluster the historical monitoring data after Z-score standardization using a K-means clustering algorithm to obtain several cluster centers, and repair each abnormal value in turn based on the nearest cluster center.

[0112] By combining Z-score anomaly detection and K-means clustering repair: first, the degree of deviation is quantified by Z-score to identify outliers, and then the cluster center value is used for repair. Compared with traditional single methods, it can better maintain the data distribution characteristics and avoid sample bias caused by simple elimination.

[0113] The step S3 is specifically as follows:

[0114] Segmenting the data set based on time segmentation labels to obtain a number of data subsets, and screening significant factors between the data subsets whose change rates exceed a preset change threshold using the Limma algorithm;

[0115] Calculating a first drought risk level for each of the significant factors using a COX survival regression model, calculating a second drought risk level for each of the significant factors using a GLM generalized linear model, cross-validating the first drought risk level and the second drought risk level to screen N environmental factors with the greatest impact on drought conditions under different time segment labels from each of the significant factors, where N is a positive integer; and constructing an environmental factor correlation graph based on the screened environmental factors and time segment labels.

[0116] The significant factors are maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, diurnal temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure, minimum air pressure, average air pressure, average daytime air pressure, average nighttime air pressure, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ global radiation, minimum TBQ global radiation, daily average TBQ global radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity, minimum soil salinity, average soil salinity, maximum soil electrical conductivity, minimum soil electrical conductivity, average soil electrical conductivity, rainfall, wind direction, wind speed or sampling location.

[0117] After screening significant factors using the Limma algorithm, the risk level was double-calculated using COX survival regression (time-related risk model) and GLM generalized linear model (nonlinear relationship modeling). Cross-validation was used to ensure the statistical significance of environmental factor screening and effectively reduce the risk of misselection of a single model.

[0118] By innovatively applying the Limma algorithm for differential gene screening in bioinformatics to the significance analysis of environmental factors, and utilizing its linear model fitting and Bayesian statistical advantages, we can efficiently identify sensitive parameters that change over time in the tea garden microenvironment (such as day and night temperature differences and CO2 concentration fluctuations), breaking through the limitations of traditional variance analysis.

[0119] By constructing an environmental factor correlation diagram through time segments (such as months), the time-varying characteristics of the relationship between environmental factors and drought conditions are captured, which is more in line with the seasonal laws of agricultural production than static correlation analysis.

[0120] In step S4, the multimodal feature extraction layer is constructed based on the meteorological time series encoding module, the soil dynamic encoding module and the spatiotemporal embedding module;

[0121] The meteorological time series encoding module is used to extract time-dependent features (light, temperature and humidity, carbon dioxide, air pressure, PM, radiation, etc.) from the monitoring data through a bidirectional LSTM network. The soil dynamic encoding module is used to extract soil dynamic features (temperature, pH, salinity, and conductivity) from the monitoring data through a fully connected network with residual connections, introducing residual connections to enhance gradient propagation. The spatiotemporal embedding module is used to extract spatiotemporal features from the monitoring data through an embedding layer (mapping longitude and latitude into high-dimensional vectors and decomposing the sampling time into periodic codes of year / month / day / time period).

[0122] The bidirectional LSTM network is used to capture the long-term and short-term temporal dependencies of meteorological factors (such as the cumulative effects of continuous droughts); the residual fully connected network is used to model the dynamic nonlinear relationships of soil parameters (such as the coupled changes in salinity and electrical conductivity); and the spatiotemporal embedding module is used to encode the joint distribution characteristics of spatial position and time (such as the impact of slope aspect on water evaporation).

[0123] The long-term dependence of meteorological parameters such as light and temperature is captured through meteorological time series coding (bidirectional LSTM), the mutation characteristics of soil indicators such as soil pH and salinity are learned through soil dynamic coding (residual FCN), and the joint features of geographic location and timestamp are fused through the spatiotemporal embedding module. That is, the parallel extraction of three types of heterogeneous features fully explores the potential correlation of data, thereby greatly improving the accuracy of tea garden drought prediction.

[0124] By setting the soil dynamic encoding module to adopt a residual fully connected network (instead of a complex CNN), the number of parameters can be reduced while avoiding gradient vanishing; combined with the GRU gated recurrent unit (lighter than LSTM), a balance between high-precision prediction and low computing resources is achieved, and it is suitable for edge computing devices (such as field embedded terminals).

[0125] The feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features by combining the environmental factor correlation graph with a multi-head attention mechanism, and output multimodal features through 1D convolutional layer fusion;

[0126] Through the multi-head attention mechanism, combined with the environmental factor correlation graph, the feature weights are dynamically adjusted, so that the model can automatically focus on key features according to different time periods and regional characteristics, thereby improving the effectiveness of feature combinations.

[0127] The soil entropy prediction layer is used to infer multimodal features through a gated recurrent unit (GRU) and a feature decoupling head to calculate soil entropy values. The feature decoupling head outputs soil parameter prediction values ​​through independent fully connected branches, achieving multi-task joint optimization.

[0128] By setting the soil moisture prediction layer to use GRU+feature decoupling head, it not only retains the gated transmission of sequence information, but also separates redundant noise through decoupling operations, which is more interpretable than the traditional end-to-end prediction structure.

[0129] The specific drought classification rules are as follows:

[0130] The soil entropy value quantile is between (75,100], corresponding to the first level of drought; the quantile is between (25,75], corresponding to the second level of drought; the quantile is between (5,25], corresponding to the third level of drought; the quantile is between (0,5], corresponding to the fourth level of drought. In specific implementation, the quantile is based on the distribution characteristics of the soil entropy value to adaptively define the level threshold (rather than a fixed threshold) to adapt to the baseline differences in different climate zones.

[0131] The step S5 is specifically as follows:

[0132] Using a stratified sampling technique, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The tea garden drought prediction model is trained using the training set. During the training process, the hyperparameters of the tea garden drought prediction model, including at least a learning rate, a batch size, and a number of training rounds, are continuously optimized until the loss function converges or a preset convergence condition is reached;

[0133] By maintaining data distribution consistency through stratified sampling at a ratio of 8:1:1, and combining automatic hyperparameter optimization (learning rate, batch size, etc.) to improve training efficiency, hardware acceleration deployment ensures real-time prediction and meets the timeliness requirements of tea garden monitoring.

[0134] The trained tea garden drought prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the verification fails, and the training set is expanded to continue training. If so, the verification passes, and:

[0135] Calculating the recall rate, F1 value and mean square error of the test set to test the tea garden drought prediction model, and deploying the tea garden drought prediction model that passes the test through hardware acceleration technology;

[0136] The step S7 is specifically as follows:

[0137] The heat map generation engine converts the tea garden drought prediction results into a drought heat map through an adaptive color scale mapping algorithm (for example, dynamically adjusting the warning threshold based on regional historical data) and a multi-scale visualization strategy (for example, displaying regional hotspots in the global view and overlaying terrain contour lines in the local view). The drought heat map is loaded onto the GIS map and displayed in real time through a visualization interface. The visualization interface refreshes the drought heat map based on a preset refresh cycle.

[0138] By adopting an adaptive color scale mapping algorithm to dynamically adjust the scale range of the thermal map, combining multi-scale visualization strategies (such as regional / plot / plant level views), and cooperating with GIS maps to achieve precise spatial positioning, it supports multi-granularity drought analysis from macro trends to micro areas.

[0139] A preferred embodiment of a visual monitoring system for drought conditions in tea gardens according to the present invention includes the following modules:

[0140] A historical monitoring data collection module is used to collect a large amount of historical monitoring data from the tea garden through a sensor array;

[0141] A data set expansion module, configured to construct a data set after preprocessing each of the historical monitoring data, and perform a sample expansion operation on the data set;

[0142] An environmental factor correlation graph generation module is used to analyze the data set using the Limma algorithm, the COX survival regression model, and the GLM generalized linear model to obtain an environmental factor correlation graph;

[0143] A tea garden drought prediction model creation module is used to create a tea garden drought prediction model based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer, and drought mapping output layer, and set the loss function of the tea garden drought prediction model;

[0144] The multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features and spatiotemporal features according to the environmental factor correlation diagram, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction result with the drought period, drought level and drought location; the drought mapping output layer outputs the probability of the drought level (level 1 drought / level 2 drought / level 3 drought / level 4 drought), the probability of the drought period, and the probability of the drought location through the fully connected layer + Softmax, and introduces the temperature scaling technology to calibrate the prediction confidence;

[0145] A tea garden drought prediction model training module is used to divide the data set into a training set, a validation set, and a test set; train the tea garden drought prediction model using the training set and a loss function; verify the trained tea garden drought prediction model using the validation set; test the verified tea garden drought prediction model using the test set; and deploy the tested tea garden drought prediction model using hardware acceleration technology; use the AdamW optimizer in the training process, combined with cosine annealing learning rate scheduling;

[0146] A tea garden drought prediction module is used to collect real-time monitoring data from the tea garden through a sensor array, pre-process the real-time monitoring data, and input the data into the deployed tea garden drought prediction model to obtain a tea garden drought prediction result;

[0147] The tea garden drought visualization display module is used to convert the tea garden drought prediction results into a drought heat map and display the drought heat map in real time through a visualization interface.

[0148] By forming a complete technical closed loop from data collection → preprocessing → model training → visualization, the synergistic effect of innovations in each link makes the overall solution significantly superior to traditional monitoring methods in key indicators such as prediction accuracy (F1 value, mean square error), result interpretability, and system response speed.

[0149] Dynamic monitoring of all elements of the tea garden environment is achieved through a multi-dimensional sensor array, and data quality is improved by combining dual anomaly repair with Z-score and K-means and autoregressive data expansion. The Limma algorithm, COX survival regression and GLM model are innovatively integrated to screen time-varying environmental factors, and a multimodal deep learning model (bidirectional LSTM, residual network, spatiotemporal embedding) is constructed to extract heterogeneous features. The features are dynamically fused based on an adaptive attention mechanism, supplemented by quantile dynamic classification rules and hardware acceleration deployment, to achieve high-precision soil moisture prediction. At the same time, through GIS integrated multi-scale heat map visualization and closed-loop iteration mechanism, macro trend analysis and micro precise positioning are taken into account, forming a full-chain technical innovation from data collection, model reasoning to decision support, significantly improving the timeliness, accuracy and adaptability of drought monitoring to agricultural scenarios, and having the dual advantages of interdisciplinary method integration and engineering implementation.

[0150] The historical monitoring data collection module is specifically used for:

[0151] A large amount of historical monitoring data is collected from the tea garden through a sensor array including humidity sensors, light intensity sensors, temperature sensors, carbon dioxide sensors, air pressure sensors, particulate matter sensors, global solar radiation sensors, pH sensors, salinity sensors, conductivity sensors, rainfall sensors, wind speed and direction sensors, and positioners;

[0152] By using an array of 13 types of sensors (such as humidity, light, CO2, etc.) to collect data, covering multi-dimensional parameters such as meteorology, soil, and environmental pollutants (such as maximum / average / daytime and nighttime sub-item data), dynamic monitoring of all elements of the tea garden microenvironment is achieved, significantly improving data comprehensiveness and monitoring accuracy.

[0153] By collecting sampling location data through locators and dynamically binding it to GIS maps, drought forecasts can be accurately mapped to spatial coordinates. Compared to traditional methods that only mark plot numbers, this technology can accurately determine longitude and latitude coordinates, providing centimeter-level positioning support for precision agriculture operations such as drone irrigation.

[0154] The historical monitoring data at least include maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, daily temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure value, minimum air pressure value, average air pressure value, average daytime air pressure value, average nighttime air pressure value, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ total radiation, minimum TBQ total radiation, daily average TBQ total radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity value, minimum soil salinity value, average soil salinity value, maximum soil conductivity, minimum soil conductivity, average soil conductivity, rainfall, wind direction, wind speed, sampling location and sampling time.

[0155] The data set expansion module is specifically used to:

[0156] After filling missing values ​​and repairing outliers for each of the historical monitoring data, each of the historical monitoring data is labeled with soil entropy, drought level, and time segment labels to complete the preprocessing of each of the historical monitoring data. A data set is constructed based on the preprocessed historical monitoring data, and a sample expansion operation is performed on the data set through an autoregressive model; the time segment label is the month of the sampling time;

[0157] By using the autoregressive model to expand the sample, we can not only maintain the statistical characteristics of the time series, but also solve the problem of insufficient seasonal samples in agricultural data and improve the generalization ability of the model.

[0158] By using months as time segment labels (rather than simple quarterly or annual divisions), it is more in line with the growth cycle of tea trees (such as the spring tea picking period, summer dormancy period, etc.), so that the results of the environmental factor correlation analysis are highly matched with the physiological stages of tea trees, thereby improving the agronomic rationality of drought prediction.

[0159] The missing value filling is based on the K-nearest neighbor filling method; the outlier repair is based on the Z-score algorithm and the K-means clustering algorithm, specifically:

[0160] Perform Z-score standardization on each of the historical monitoring data, calculate the Z-score value of each of the historical monitoring data after Z-score standardization, and compare the Z-score value with a preset abnormality threshold to screen out abnormal values; cluster the historical monitoring data after Z-score standardization using a K-means clustering algorithm to obtain several cluster centers, and repair each abnormal value in turn based on the nearest cluster center.

[0161] By combining Z-score anomaly detection and K-means clustering repair: first, the degree of deviation is quantified by Z-score to identify outliers, and then the cluster center value is used for repair. Compared with traditional single methods, it can better maintain the data distribution characteristics and avoid sample bias caused by simple elimination.

[0162] The environmental factor correlation graph generation module is specifically used for:

[0163] Segmenting the data set based on time segmentation labels to obtain a number of data subsets, and screening significant factors between the data subsets whose change rates exceed a preset change threshold using the Limma algorithm;

[0164] Calculating a first drought risk level for each of the significant factors using a COX survival regression model, calculating a second drought risk level for each of the significant factors using a GLM generalized linear model, cross-validating the first drought risk level and the second drought risk level to screen N environmental factors with the greatest impact on drought conditions under different time segment labels from each of the significant factors, where N is a positive integer; and constructing an environmental factor correlation graph based on the screened environmental factors and time segment labels.

[0165] The significant factors are maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, diurnal temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure, minimum air pressure, average air pressure, average daytime air pressure, average nighttime air pressure, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ global radiation, minimum TBQ global radiation, daily average TBQ global radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity, minimum soil salinity, average soil salinity, maximum soil electrical conductivity, minimum soil electrical conductivity, average soil electrical conductivity, rainfall, wind direction, wind speed or sampling location.

[0166] After screening significant factors using the Limma algorithm, the risk level was double-calculated using COX survival regression (time-related risk model) and GLM generalized linear model (nonlinear relationship modeling). Cross-validation was used to ensure the statistical significance of environmental factor screening and effectively reduce the risk of misselection of a single model.

[0167] By innovatively applying the Limma algorithm for differential gene screening in bioinformatics to the significance analysis of environmental factors, and utilizing its linear model fitting and Bayesian statistical advantages, we can efficiently identify sensitive parameters that change over time in the tea garden microenvironment (such as day and night temperature differences and CO2 concentration fluctuations), breaking through the limitations of traditional variance analysis.

[0168] By constructing an environmental factor correlation diagram through time segments (such as months), the time-varying characteristics of the relationship between environmental factors and drought conditions are captured, which is more in line with the seasonal laws of agricultural production than static correlation analysis.

[0169] In the tea garden drought prediction model creation module, the multimodal feature extraction layer is constructed based on the meteorological time series coding module, the soil dynamic coding module and the spatiotemporal embedding module;

[0170] The meteorological time series encoding module is used to extract time-dependent features (light, temperature and humidity, carbon dioxide, air pressure, PM, radiation, etc.) from the monitoring data through a bidirectional LSTM network. The soil dynamic encoding module is used to extract soil dynamic features (temperature, pH, salinity, and conductivity) from the monitoring data through a fully connected network with residual connections, introducing residual connections to enhance gradient propagation. The spatiotemporal embedding module is used to extract spatiotemporal features from the monitoring data through an embedding layer (mapping longitude and latitude into high-dimensional vectors and decomposing the sampling time into periodic codes of year / month / day / time period).

[0171] The bidirectional LSTM network is used to capture the long-term and short-term temporal dependencies of meteorological factors (such as the cumulative effects of continuous droughts); the residual fully connected network is used to model the dynamic nonlinear relationships of soil parameters (such as the coupled changes in salinity and electrical conductivity); and the spatiotemporal embedding module is used to encode the joint distribution characteristics of spatial position and time (such as the impact of slope aspect on water evaporation).

[0172] The long-term dependence of meteorological parameters such as light and temperature is captured through meteorological time series coding (bidirectional LSTM), the mutation characteristics of soil indicators such as soil pH and salinity are learned through soil dynamic coding (residual FCN), and the joint features of geographic location and timestamp are fused through the spatiotemporal embedding module. That is, the parallel extraction of three types of heterogeneous features fully explores the potential correlation of data, thereby greatly improving the accuracy of tea garden drought prediction.

[0173] By setting the soil dynamic encoding module to adopt a residual fully connected network (instead of a complex CNN), the number of parameters can be reduced while avoiding gradient vanishing; combined with the GRU gated recurrent unit (lighter than LSTM), a balance between high-precision prediction and low computing resources is achieved, and it is suitable for edge computing devices (such as field embedded terminals).

[0174] The feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features by combining the environmental factor correlation graph with a multi-head attention mechanism, and output multimodal features through 1D convolutional layer fusion;

[0175] Through the multi-head attention mechanism, combined with the environmental factor correlation graph, the feature weights are dynamically adjusted, so that the model can automatically focus on key features according to different time periods and regional characteristics, thereby improving the effectiveness of feature combinations.

[0176] The soil entropy prediction layer is used to infer multimodal features through a gated recurrent unit (GRU) and a feature decoupling head to calculate soil entropy values. The feature decoupling head outputs soil parameter prediction values ​​through independent fully connected branches, achieving multi-task joint optimization.

[0177] By setting the soil moisture prediction layer to use GRU+feature decoupling head, it not only retains the gated transmission of sequence information, but also separates redundant noise through decoupling operations, which is more interpretable than the traditional end-to-end prediction structure.

[0178] The specific drought classification rules are as follows:

[0179] The soil entropy value quantile is between (75,100], corresponding to the first level of drought; the quantile is between (25,75], corresponding to the second level of drought; the quantile is between (5,25], corresponding to the third level of drought; the quantile is between (0,5], corresponding to the fourth level of drought. In specific implementation, the quantile is based on the distribution characteristics of the soil entropy value to adaptively define the level threshold (rather than a fixed threshold) to adapt to the baseline differences in different climate zones.

[0180] The tea garden drought prediction model training module is specifically used for:

[0181] Using a stratified sampling technique, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The tea garden drought prediction model is trained using the training set. During the training process, the hyperparameters of the tea garden drought prediction model, including at least a learning rate, a batch size, and a number of training rounds, are continuously optimized until the loss function converges or a preset convergence condition is reached;

[0182] By maintaining data distribution consistency through stratified sampling at a ratio of 8:1:1, and combining automatic hyperparameter optimization (learning rate, batch size, etc.) to improve training efficiency, hardware acceleration deployment ensures real-time prediction and meets the timeliness requirements of tea garden monitoring.

[0183] The trained tea garden drought prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the verification fails, and the training set is expanded to continue training. If so, the verification passes, and:

[0184] Calculating the recall rate, F1 value and mean square error of the test set to test the tea garden drought prediction model, and deploying the tea garden drought prediction model that passes the test through hardware acceleration technology;

[0185] The tea garden drought visualization display module is specifically used for:

[0186] The heat map generation engine converts the tea garden drought prediction results into a drought heat map through an adaptive color scale mapping algorithm (for example, dynamically adjusting the warning threshold based on regional historical data) and a multi-scale visualization strategy (for example, displaying regional hotspots in the global view and overlaying terrain contour lines in the local view). The drought heat map is loaded onto the GIS map and displayed in real time through a visualization interface. The visualization interface refreshes the drought heat map based on a preset refresh cycle.

[0187] By adopting an adaptive color scale mapping algorithm to dynamically adjust the scale range of the thermal map, combining multi-scale visualization strategies (such as regional / plot / plant level views), and cooperating with GIS maps to achieve precise spatial positioning, it supports multi-granularity drought analysis from macro trends to micro areas.

[0188] In summary, the advantages of the present invention are:

[0189] 1. A large amount of historical monitoring data is collected from the tea garden through the sensor array to construct a data set, and a sample expansion operation is performed on the data set; then the data set is analyzed by the Limma algorithm, COX survival regression model and GLM generalized linear model to obtain an environmental factor correlation map; then a tea garden drought prediction model is created based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer and drought mapping output layer, and the loss function of the tea garden drought prediction model is set; the multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features and spatiotemporal features according to the environmental factor correlation map, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction results with drought period, drought level and drought location; then the data set is divided into training set, validation set and test set. A test set is used to train, verify, and test the tea garden drought prediction model. The tea garden drought prediction model that passes the test is deployed using hardware acceleration technology. Then, real-time monitoring data is collected from the tea garden through a sensor array. After preprocessing, the real-time monitoring data is input into the deployed tea garden drought prediction model to obtain the tea garden drought prediction results. The tea garden drought prediction results are converted into drought heat maps for real-time display through a visualization interface. That is, the tea garden drought prediction model is used to predict the tea garden drought. Before the prediction, the feature fusion layer of the tea garden drought prediction model calls the environmental factor correlation map to adjust the weight of each feature. That is, the weight of the features corresponding to the N environmental factors with the greatest impact on drought prediction under the current time segment label is increased. The tea garden drought prediction results are output in combination with the drought classification rules to facilitate refined irrigation of the tea garden. The tea garden drought prediction results are converted into drought heat maps for visualization, which can accurately judge the spatial distribution of drought, ultimately greatly improving the accuracy of tea garden drought prediction, the refinement of irrigation decision-making, and the intuitiveness of drought display.

[0190] 2. By using an array of 13 types of sensors (such as humidity, light, CO2, etc.) to collect data, covering multi-dimensional parameters such as meteorology, soil, and environmental pollutants (such as maximum / average / daytime and nighttime sub-item data), dynamic monitoring of all elements of the tea garden microenvironment is achieved, significantly improving data comprehensiveness and monitoring accuracy.

[0191] 3. By combining Z-score anomaly detection and K-means clustering repair: First, identify outliers by quantifying the degree of deviation through Z-score, and then use the cluster center value to repair them. Compared with traditional single methods, this method can better maintain the data distribution characteristics and avoid sample bias caused by simple elimination.

[0192] 4. By using the autoregressive model for sample expansion, the statistical characteristics of the time series are maintained, the problem of insufficient seasonal samples in agricultural data is solved, and the generalization ability of the model is improved.

[0193] 5. After screening significant factors using the Limma algorithm, the risk level is calculated using both COX survival regression (time-related risk model) and GLM generalized linear model (nonlinear relationship modeling). Cross-validation is used to ensure the statistical significance of environmental factor screening and effectively reduce the risk of misselection of a single model.

[0194] 6. Constructing an environmental factor correlation diagram by time segmentation (such as month) can capture the time-varying characteristics of the relationship between environmental factors and drought conditions. Compared with static correlation analysis, it is more in line with the seasonal laws of agricultural production.

[0195] 7. Meteorological time series coding (bidirectional LSTM) is used to capture the long-term dependence of meteorological parameters such as light and temperature. Soil dynamic coding (residual FCN) is used to learn the mutation characteristics of indicators such as soil pH and salinity. The spatiotemporal embedding module fuses the joint features of geographic location and timestamp. In other words, the parallel extraction of three types of heterogeneous features fully explores the potential correlation of the data, thereby greatly improving the accuracy of tea garden drought prediction.

[0196] 8. Through the multi-head attention mechanism, combined with the environmental factor correlation graph, the feature weights are dynamically adjusted, so that the model can automatically focus on key features according to different time periods and regional characteristics, thereby improving the effectiveness of feature combinations.

[0197] 9. By setting the soil moisture prediction layer to use GRU+feature decoupling head, it not only retains the gated transmission of sequence information, but also separates redundant noise through decoupling operations, which is more interpretable than the traditional end-to-end prediction structure.

[0198] 10. Maintaining data distribution consistency through stratified sampling at a ratio of 8:1:1, improving training efficiency through automatic optimization of hyperparameters (learning rate, batch size, etc.), and ensuring real-time prediction through hardware acceleration deployment to meet the timeliness requirements of tea garden monitoring.

[0199] 11. By adopting an adaptive color scale mapping algorithm to dynamically adjust the scale range of the thermal map, combining multi-scale visualization strategies (such as regional / plot / plant level views), and cooperating with GIS maps to achieve precise spatial positioning, it supports multi-granularity drought analysis from macro trends to micro areas.

[0200] 12. By forming a complete technical closed loop from data collection → preprocessing → model training → visualization, the synergistic effect of innovations in each link makes the overall solution significantly superior to traditional monitoring methods in key indicators such as prediction accuracy (F1 value, mean square error), result interpretability, and system response speed.

[0201] 13. Locators collect sampling location data and dynamically bind it to GIS maps, enabling precise spatial mapping of drought forecast results. Compared to traditional methods that simply mark plot numbers, this technology can accurately map latitude and longitude coordinates, providing centimeter-level positioning support for precision agriculture operations such as drone irrigation.

[0202] 14. By using months as time segment labels (rather than simple quarterly or annual divisions), it is more in line with the growth cycle of tea trees (such as the spring tea picking period, summer dormancy period, etc.), so that the results of the environmental factor correlation analysis are highly matched with the physiological stages of tea trees, thereby improving the agronomic rationality of drought prediction.

[0203] 15. By innovatively applying the Limma algorithm for differential gene screening in bioinformatics to the significance analysis of environmental factors, and taking advantage of its linear model fitting and Bayesian statistical advantages, we can efficiently identify sensitive parameters that change over time in the tea garden microenvironment (such as day and night temperature differences and CO2 concentration fluctuations), breaking through the limitations of traditional variance analysis.

[0204] 16. By setting the soil dynamic encoding module to adopt a residual fully connected network (instead of a complex CNN), the number of parameters can be reduced while avoiding gradient vanishing; combined with the GRU gated recurrent unit (lighter than LSTM), a balance between high-precision prediction and low computing resources is achieved, and it is suitable for edge computing devices (such as field embedded terminals).

[0205] 17. Dynamic monitoring of all elements of the tea garden environment is achieved through a multi-dimensional sensor array, and data quality is improved by combining dual anomaly repair with Z-score and K-means and autoregressive data expansion; the Limma algorithm, COX survival regression and GLM model are innovatively integrated to screen time-varying environmental factors, and a multimodal deep learning model (bidirectional LSTM, residual network, spatiotemporal embedding) is constructed to extract heterogeneous features. The features are dynamically fused based on the adaptive attention mechanism, supplemented by quantile dynamic classification rules and hardware acceleration deployment, to achieve high-precision soil moisture prediction; at the same time, through the multi-scale heat map visualization and closed-loop iteration mechanism integrated by GIS, taking into account both macro trend analysis and micro precise positioning, a full-chain technological innovation is formed from data collection, model reasoning to decision support, which significantly improves the timeliness, accuracy and adaptability of drought monitoring to agricultural scenarios, and has the dual advantages of interdisciplinary method integration and engineering implementation.

[0206] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for visually monitoring drought conditions in a tea garden, characterized by: The steps include: Step S1: collecting a large amount of historical monitoring data from the tea garden through a sensor array; Step S2: constructing a data set after preprocessing each of the historical monitoring data, and performing a sample expansion operation on the data set; Step S3, analyzing the data set using the Limma algorithm, the COX survival regression model, and the GLM generalized linear model to obtain an environmental factor correlation graph; Step S4: creating a tea garden drought prediction model based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer, and drought mapping output layer, and setting a loss function of the tea garden drought prediction model; The multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features, and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features according to the environmental factor correlation diagram, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction result with the drought period, drought level, and drought location; Step S5, dividing the data set into a training set, a validation set, and a test set, training the tea garden drought prediction model using the training set and a loss function, validating the trained tea garden drought prediction model using the validation set, testing the verified tea garden drought prediction model using the test set, and deploying the tested tea garden drought prediction model using hardware acceleration technology; Step S6: collecting real-time monitoring data from the tea garden through the sensor array, pre-processing the real-time monitoring data, and inputting the data into the deployed tea garden drought prediction model to obtain a tea garden drought prediction result; Step S7: converting the tea garden drought prediction result into a drought heat map, and displaying the drought heat map in real time through a visual interface.

2. A method for visually monitoring drought conditions in a tea garden according to claim 1, characterized in that: The step S1 is specifically as follows: A large amount of historical monitoring data is collected from the tea garden through a sensor array including humidity sensors, light intensity sensors, temperature sensors, carbon dioxide sensors, air pressure sensors, particulate matter sensors, global solar radiation sensors, pH sensors, salinity sensors, conductivity sensors, rainfall sensors, wind speed and direction sensors, and positioners; The historical monitoring data at least include maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, daily temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure value, minimum air pressure value, average air pressure value, average daytime air pressure value, average nighttime air pressure value, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ total radiation, minimum TBQ total radiation, daily average TBQ total radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity value, minimum soil salinity value, average soil salinity value, maximum soil conductivity, minimum soil conductivity, average soil conductivity, rainfall, wind direction, wind speed, sampling location and sampling time.

3. The method for visually monitoring drought conditions in a tea garden according to claim 1, wherein: The step S2 is specifically as follows: After filling missing values ​​and repairing outliers for each of the historical monitoring data, each of the historical monitoring data is labeled with soil entropy, drought level, and time segment labels to complete the preprocessing of each of the historical monitoring data. A data set is constructed based on the preprocessed historical monitoring data, and a sample expansion operation is performed on the data set through an autoregressive model; the time segment label is the month of the sampling time; The missing value filling is based on the K-nearest neighbor filling method; the outlier repair is based on the Z-score algorithm and the K-means clustering algorithm, specifically: Perform Z-score standardization on each of the historical monitoring data, calculate the Z-score value of each of the historical monitoring data after Z-score standardization, and compare the Z-score value with a preset abnormality threshold to screen out abnormal values; cluster the historical monitoring data after Z-score standardization using a K-means clustering algorithm to obtain several cluster centers, and repair each abnormal value in turn based on the nearest cluster center.

4. The method for visually monitoring drought conditions in a tea garden according to claim 1, wherein: The step S3 is specifically as follows: Segmenting the data set based on time segmentation labels to obtain a number of data subsets, and screening significant factors between the data subsets whose change rates exceed a preset change threshold using the Limma algorithm; Calculating a first drought risk level for each of the significant factors using a COX survival regression model, calculating a second drought risk level for each of the significant factors using a GLM generalized linear model, cross-validating the first drought risk level and the second drought risk level to screen N environmental factors with the greatest impact on drought conditions under different time segment labels from each of the significant factors, where N is a positive integer; and constructing an environmental factor correlation graph based on the screened environmental factors and time segment labels. In step S4, the multimodal feature extraction layer is constructed based on the meteorological time series encoding module, the soil dynamic encoding module and the spatiotemporal embedding module; The meteorological time series encoding module is used to extract time-dependent features from the monitoring data through a bidirectional LSTM network; the soil dynamic encoding module is used to extract soil dynamic features from the monitoring data through a fully connected network with residual connections; the spatiotemporal embedding module is used to extract spatiotemporal features from the monitoring data through an embedding layer; The feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features by combining the environmental factor correlation graph with a multi-head attention mechanism, and output multimodal features through 1D convolutional layer fusion; The soil entropy prediction layer is used to infer multimodal features through gated recurrent units and feature decoupling heads to calculate soil entropy values; The specific drought classification rules are as follows: The quantile of soil entropy value is (75,100], which corresponds to the first level of drought; the quantile is (25,75], which corresponds to the second level of drought; the quantile is (5,25], which corresponds to the third level of drought; the quantile is (0,5], which corresponds to the fourth level of drought.

5. The method for visually monitoring drought conditions in a tea garden according to claim 1, wherein: The step S5 is specifically as follows: Using a stratified sampling technique, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:

1. The tea garden drought prediction model is trained using the training set. During the training process, the hyperparameters of the tea garden drought prediction model, including at least a learning rate, a batch size, and a number of training rounds, are continuously optimized until the loss function converges or a preset convergence condition is reached; The trained tea garden drought prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the verification fails, and the training set is expanded to continue training. If so, the verification passes, and: Calculating the recall rate, F1 value and mean square error of the test set to test the tea garden drought prediction model, and deploying the tea garden drought prediction model that passes the test through hardware acceleration technology; The step S7 is specifically as follows: The heat map generation engine converts the tea garden drought prediction results into a drought heat map through an adaptive color scale mapping algorithm and a multi-scale visualization strategy, loads the drought heat map onto a GIS map, and displays it in real time through a visualization interface. The visualization interface refreshes the drought heat map based on a preset refresh cycle.

6. A visual monitoring system for drought conditions in tea gardens, characterized by: Includes the following modules: A historical monitoring data collection module is used to collect a large amount of historical monitoring data from the tea garden through a sensor array; A data set expansion module, configured to construct a data set after preprocessing each of the historical monitoring data, and perform a sample expansion operation on the data set; An environmental factor correlation graph generation module is used to analyze the data set using the Limma algorithm, the COX survival regression model, and the GLM generalized linear model to obtain an environmental factor correlation graph; A tea garden drought prediction model creation module is used to create a tea garden drought prediction model based on the sequentially connected multimodal feature extraction layer, feature fusion layer, soil entropy prediction layer, and drought mapping output layer, and set the loss function of the tea garden drought prediction model; The multimodal feature extraction layer is used to extract time-dependent features, soil dynamic features, and spatiotemporal features from the monitoring data; the feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features according to the environmental factor correlation diagram, and fuse them to obtain multimodal features; the soil entropy prediction layer is used to calculate the soil entropy value based on the multimodal features; the drought mapping output layer is used to map the soil entropy value according to the set drought classification rules, and output the tea garden drought prediction result with the drought period, drought level, and drought location; A tea garden drought prediction model training module is used to divide the data set into a training set, a validation set, and a test set, train the tea garden drought prediction model using the training set and a loss function, verify the trained tea garden drought prediction model using the validation set, test the verified tea garden drought prediction model using the test set, and deploy the tested tea garden drought prediction model using hardware acceleration technology; A tea garden drought prediction module is used to collect real-time monitoring data from the tea garden through a sensor array, pre-process the real-time monitoring data, and input the data into the deployed tea garden drought prediction model to obtain a tea garden drought prediction result; The tea garden drought visualization display module is used to convert the tea garden drought prediction results into a drought heat map and display the drought heat map in real time through a visualization interface.

7. A tea garden drought visualization monitoring system according to claim 6, characterized in that: The historical monitoring data collection module is specifically used for: A large amount of historical monitoring data is collected from the tea garden through a sensor array including humidity sensors, light intensity sensors, temperature sensors, carbon dioxide sensors, air pressure sensors, particulate matter sensors, global solar radiation sensors, pH sensors, salinity sensors, conductivity sensors, rainfall sensors, wind speed and direction sensors, and positioners; The historical monitoring data at least include maximum light intensity, average light intensity, sunshine duration, maximum temperature, minimum temperature, average temperature, average daytime temperature, average nighttime temperature, daily temperature difference, maximum air humidity, minimum air humidity, average relative humidity, average daytime air humidity, average nighttime air humidity, maximum carbon dioxide concentration, minimum carbon dioxide concentration, average carbon dioxide concentration, average daytime carbon dioxide concentration, average nighttime carbon dioxide concentration, maximum air pressure value, minimum air pressure value, average air pressure value, average daytime air pressure value, average nighttime air pressure value, air pressure difference, daily average PM2.5, daily average PM10, maximum TBQ total radiation, minimum TBQ total radiation, daily average TBQ total radiation, maximum soil temperature, minimum soil temperature, daily average soil temperature, soil temperature difference, maximum soil pH, minimum soil pH, average soil pH, maximum soil salinity value, minimum soil salinity value, average soil salinity value, maximum soil conductivity, minimum soil conductivity, average soil conductivity, rainfall, wind direction, wind speed, sampling location and sampling time.

8. The tea garden drought visualization monitoring system according to claim 6, characterized in that: The data set expansion module is specifically used to: After filling missing values ​​and repairing outliers for each of the historical monitoring data, each of the historical monitoring data is labeled with soil entropy, drought level, and time segment labels to complete the preprocessing of each of the historical monitoring data. A data set is constructed based on the preprocessed historical monitoring data, and a sample expansion operation is performed on the data set through an autoregressive model; the time segment label is the month of the sampling time; The missing value filling is based on the K-nearest neighbor filling method; the outlier repair is based on the Z-score algorithm and the K-means clustering algorithm, specifically: Perform Z-score standardization on each of the historical monitoring data, calculate the Z-score value of each of the historical monitoring data after Z-score standardization, and compare the Z-score value with a preset abnormality threshold to screen out abnormal values; cluster the historical monitoring data after Z-score standardization using a K-means clustering algorithm to obtain several cluster centers, and repair each abnormal value in turn based on the nearest cluster center.

9. The tea garden drought visualization monitoring system according to claim 6, characterized in that: The environmental factor correlation graph generation module is specifically used for: Segmenting the data set based on time segmentation labels to obtain a number of data subsets, and screening significant factors between the data subsets whose change rates exceed a preset change threshold using the Limma algorithm; Calculating a first drought risk level for each of the significant factors using a COX survival regression model, calculating a second drought risk level for each of the significant factors using a GLM generalized linear model, cross-validating the first drought risk level and the second drought risk level to screen N environmental factors with the greatest impact on drought conditions under different time segment labels from each of the significant factors, where N is a positive integer; and constructing an environmental factor correlation graph based on the screened environmental factors and time segment labels. In the tea garden drought prediction model creation module, the multimodal feature extraction layer is constructed based on the meteorological time series coding module, the soil dynamic coding module and the spatiotemporal embedding module; The meteorological time series encoding module is used to extract time-dependent features from the monitoring data through a bidirectional LSTM network; the soil dynamic encoding module is used to extract soil dynamic features from the monitoring data through a fully connected network with residual connections; the spatiotemporal embedding module is used to extract spatiotemporal features from the monitoring data through an embedding layer; The feature fusion layer is used to adjust the weights of time-dependent features, soil dynamic features, and spatiotemporal features by combining the environmental factor correlation graph with a multi-head attention mechanism, and output multimodal features through 1D convolutional layer fusion; The soil entropy prediction layer is used to infer multimodal features through gated recurrent units and feature decoupling heads to calculate soil entropy values; The specific drought classification rules are as follows: The quantile of soil entropy value is (75,100], which corresponds to the first level of drought; the quantile is (25,75], which corresponds to the second level of drought; the quantile is (5,25], which corresponds to the third level of drought; the quantile is (0,5], which corresponds to the fourth level of drought.

10. The tea garden drought visualization monitoring system according to claim 6, characterized in that: The tea garden drought prediction model training module is specifically used for: Using a stratified sampling technique, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:

1. The tea garden drought prediction model is trained using the training set. During the training process, the hyperparameters of the tea garden drought prediction model, including at least a learning rate, a batch size, and a number of training rounds, are continuously optimized until the loss function converges or a preset convergence condition is reached; The trained tea garden drought prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the verification fails, and the training set is expanded to continue training. If so, the verification passes, and: Calculating the recall rate, F1 value and mean square error of the test set to test the tea garden drought prediction model, and deploying the tea garden drought prediction model that passes the test through hardware acceleration technology; The tea garden drought visualization display module is specifically used for: The heat map generation engine converts the tea garden drought prediction results into a drought heat map through an adaptive color scale mapping algorithm and a multi-scale visualization strategy, loads the drought heat map onto a GIS map, and displays it in real time through a visualization interface. The visualization interface refreshes the drought heat map based on a preset refresh cycle.

Images