An automatic fertilization and watering control method and system for landscape nursery stock

Abstract

The present invention relates to the technical field of fertilization and watering control, and particularly relates to an automatic fertilization and watering control method and system for landscape nursery stock. The method includes the following steps: collecting and analyzing the morphological data of landscape nursery stock and the soil state characteristics through electronic monitoring and soil monitoring sensors; then, predicting the density of the nursery stock roots based on the morphological data of the nursery stock, monitoring abnormal fluctuations in soil salt content, and estimating the limited root absorption situation; next, matching the dosage of slow-release fertilizer according to the limited root absorption data and abnormal fluctuations in soil salt content, and performing segmented drip irrigation ratio fusion; finally, constructing a fertilization and watering control model based on the policy gradient algorithm, and implementing automatic fertilization and watering control through a terminal. The present invention makes the fertilization and watering control technology more perfect through the optimization of the fertilization and watering control technology.

Description

Technical Field

[0001] The present invention relates to the technical field of fertilization and watering control, and particularly to an automatic fertilization and watering control method and system for landscape nursery stock. Background Art

[0002] By monitoring the morphology of nursery stock, soil characteristics, and environmental changes in real time, more accurate data can be obtained to help achieve precise fertilization and watering control. Especially with the support of multi-dimensional data such as soil humidity, temperature, and salt content, human operation errors can be effectively avoided, resource waste can be reduced, and at the same time, the scientificity and stability of plant growth can be improved. Factors such as the root growth of nursery stock, the water and nutrient supply of the soil, etc. will affect the health status of plants, and the previous manual judgment and regular fertilization and watering cannot respond to the real-time needs of plants in a timely manner. Through an automatic control system, combined with the monitoring of the soil, air, and plant conditions by intelligent sensors, the fertilization and watering amounts can be accurately adjusted according to the actual needs of the nursery stock. For example, abnormal fluctuations in soil salt content may affect the root absorption ability, and the intelligent system can adjust the fertilizer dosage according to these changes to avoid soil salt accumulation caused by excessive fertilization, thereby achieving the purpose of precise regulation. In addition, the intelligent control system can achieve segmented drip irrigation and fertilizer application to meet the needs of different plants and maximize the maintenance efficiency and quality. However, there is a problem with the traditional automatic fertilization and watering control method for landscape nursery stock that the analysis of the limited root nutrient absorption caused by soil salinization is inaccurate, resulting in large errors in automatic fertilization and watering. Summary of the Invention

[0003] Based on this, it is necessary to provide an automatic fertilization and watering control method and system for landscape nursery stock to solve at least one of the above technical problems.

[0004] To achieve the above object, an automatic fertilization and watering control method for landscape nursery stock, the method includes the following steps:

[0005] Step S1: Collect nursery stock morphology data of landscape nursery stock through electronic monitoring to obtain nursery stock morphology data; deploy soil monitoring sensors under each landscape nursery stock, and monitor the soil state characteristics through the soil detection sensors to obtain soil state characteristic monitoring feature data;

[0006] Step S2: Predict the root density of the nursery stock based on the nursery stock morphology data to generate nursery stock root density prediction data; analyze the abnormal fluctuations in soil salt content of the soil state characteristic monitoring feature data to obtain soil salt content abnormal fluctuation data; estimate the limited root absorption based on the soil salt content abnormal fluctuation data for the nursery stock root density prediction data to obtain limited root absorption estimation data;

[0007] Step S3: According to the root absorption limitation estimation data, perform slow-release fertilizer dosage control matching on the abnormal soil salt content fluctuation data to obtain slow-release fertilizer dosage control data; based on the slow-release fertilizer dosage control data, perform segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio for fertilization and watering.

[0008] Step S4: Based on the policy gradient algorithm, construct a fertilization and watering control model for the segmented drip irrigation ratio of fertilization and watering to obtain a garden nursery stock fertilization and watering control model; send the garden nursery stock fertilization and watering control model to the terminal to execute the automatic fertilization and watering control of landscape nursery stock.

[0009] Preferably, step S1 includes the following steps:

[0010] Step S11: Collect nursery stock morphological data of landscape nursery stock through electronic monitoring to obtain nursery stock morphological data.

[0011] Step S12: Deploy soil monitoring sensors under each landscape nursery stock, and monitor the soil state characteristics through the soil detection sensors to obtain soil state characteristic monitoring data.

[0012] Step S13: Clean the soil state characteristic monitoring data to obtain the cleaned soil state characteristic monitoring data.

[0013] Step S14: Analyze the state characteristics of the cleaned soil state characteristic monitoring data to obtain the soil state characteristic monitoring characteristic data.

[0014] Preferably, step S2 includes the following steps:

[0015] Step S21: Predict the density of nursery stock roots based on the nursery stock morphological data to generate nursery stock root density prediction data.

[0016] Step S22: Analyze the abnormal soil salt content fluctuations of the soil state characteristic monitoring characteristic data to obtain abnormal soil salt content fluctuation data.

[0017] Step S23: Based on the abnormal soil salt content fluctuation data, analyze the loss of soil nutrient imbalance effectiveness of the soil state characteristic monitoring characteristic data to obtain soil nutrient imbalance effectiveness loss data.

[0018] Step S24: Estimate the root absorption limitation of the nursery stock root density prediction data according to the abnormal soil salt content fluctuation data and the soil nutrient imbalance effectiveness loss data to obtain root absorption limitation estimation data.

[0019] Preferably, step S23 includes the following steps:

[0020] Step S231: Perform ion concentration decomposition on the abnormal fluctuation data of soil salt content to generate time-series change data of salt-containing ion concentration;

[0021] Step S232: Extract the fluctuation of organic matter content from the monitoring feature data of soil state characteristics to obtain the fluctuation data of soil organic matter content;

[0022] Step S233: Conduct non-linear strength regression analysis of cell membrane damage based on the time-series change data of salt-containing ion concentration to obtain the regression data of cell membrane damage strength;

[0023] Step S234: Perform an equivalent simulation prediction of organic matter nutrient leakage on the fluctuation data of soil organic matter content according to the regression data of cell membrane damage strength to obtain the equivalent prediction data of nutrient leakage;

[0024] Step S235: Analyze the increment of ion competition intensity for the fluctuation data of soil organic matter content according to the time-series change data of salt-containing ion concentration to generate ion competition intensity increment data;

[0025] Step S236: Conduct an analysis of the effective loss of soil nutrient imbalance based on the equivalent prediction data of nutrient leakage and the ion competition intensity increment data to obtain the effective loss data of soil nutrient imbalance.

[0026] Preferably, step S235 includes the following steps:

[0027] Perform a mutation increment index calculation on the time-series change data of salt-containing ion concentration to obtain the mutation increment index of salt-containing ion concentration;

[0028] Based on the mutation increment index of salt-containing ion concentration, perform a continuous concentration mutation approximation combination on the time-series change data of salt-containing ion concentration to generate concentration continuous mutation approximation combination data;

[0029] Based on the concentration continuous mutation approximation combination data, perform a preemptive continuous intensity simulation of salt-containing ion absorption sites on the fluctuation data of soil organic matter content to obtain the preemptive continuous intensity data of salt-containing ion absorption sites;

[0030] Perform an ion flux time-series interval difference calculation on the preemptive continuous intensity data of absorption sites to obtain the salt-containing ion flux difference at absorption sites;

[0031] Analyze the increment of ion competition intensity according to the salt-containing ion flux difference at absorption sites to generate ion competition intensity increment data.

[0032] Preferably, step S24 includes the following steps:

[0033] Step S241: Perform a fluctuation amplitude deviation rate operation on the abnormal fluctuation data of soil salt content to obtain salt content fluctuation amplitude difference data;

[0034] Step S242: Perform quantization conversion of the increased ratio of the soil solution osmotic pressure based on the difference data of the salt content fluctuation amplitude to obtain the increased ratio of the soil solution osmotic pressure;

[0035] Step S243: Perform integral of the fluctuation of the microbial nutrient conversion inhibition value based on the increased ratio of the soil solution osmotic pressure to obtain the fluctuation value of the nutrient conversion inhibition;

[0036] Step S244: Estimate the cumulative gradient of soil compaction based on the difference data of the salt content fluctuation amplitude and the increased ratio of the soil solution osmotic pressure to obtain the cumulative gradient data of soil compaction;

[0037] Step S245: Evaluate the restriction gradient of the root growth / respiration based on the cumulative gradient data of soil compaction for the predicted data of the density of the seedling roots to obtain the restriction gradient data of the root growth / respiration;

[0038] Step S246: Estimate the root absorption limitation for the predicted data of the density of the seedling roots according to the fluctuation value of the nutrient conversion inhibition, the restriction gradient data of the root growth / respiration, and the effective loss data of soil nutrient imbalance to obtain the estimated data of root absorption limitation.

[0039] Preferably, step S3 includes the following steps:

[0040] Step S31: Control and match the application rate of slow-release fertilizer for the abnormal fluctuation data of the soil salt content according to the estimated data of root absorption limitation to obtain the control data of the slow-release fertilizer application rate;

[0041] Step S32: Learn the application rate control logic for the control data of the slow-release fertilizer application rate to obtain the learning data of the slow-release fertilizer application rate control;

[0042] Step S33: Perform segmented drip irrigation ratio fusion based on the learning data of the slow-release fertilizer application rate control and the estimated data of root absorption limitation to obtain the segmented drip irrigation ratio of fertilization and watering.

[0043] Preferably, step S33 includes the following steps:

[0044] Step S331: Deduce the root respiration conversion demand space based on the estimated data of root absorption limitation to obtain the root respiration conversion demand space;

[0045] Step S332: Calculate the nutrient absorption rate interval per unit time for the estimated data of root absorption limitation to obtain the nutrient absorption rate interval of the roots per unit time;

[0046] Step S333: Analyze the fertilizer efficiency release rate for the learning data of the slow-release fertilizer application rate control to obtain the fertilizer efficiency application release rate;

[0047] Step S334: Perform the dosage absorption extreme point segmentation processing on the fertilizer efficiency dosage release rate per unit time according to the root respiration conversion demand space and the root nutrient absorption rate range to obtain the root fertilizer efficiency release and absorption extreme value data;

[0048] Step S335: Analyze the absorption and release load ratio of the root fertilizer efficiency release and absorption extreme value data to obtain the root fertilizer efficiency absorption and release load ratio;

[0049] Step S336: Perform segmented drip irrigation ratio fusion based on the root fertilizer efficiency absorption and release load ratio, the root nutrient absorption rate range, and the root respiration conversion demand space to obtain the segmented drip irrigation ratio for fertilization and watering.

[0050] Preferably, step S4 includes the following steps:

[0051] Step S41: Normalize the segmented drip irrigation ratio for fertilization and watering to obtain the normalized segmented drip irrigation ratio for fertilization and watering;

[0052] Step S42: Construct a fertilization and watering control model for the segmented drip irrigation ratio for fertilization and watering based on the policy gradient algorithm to obtain a fertilization and watering control model for landscape nursery stock;

[0053] Step S43: Send the fertilization and watering control model for landscape nursery stock to the terminal to perform automatic fertilization and watering control for landscape nursery stock.

[0054] Preferably, the present invention also provides an automatic fertilization and watering control system for landscape nursery stock, which is used to perform the automatic fertilization and watering control method for landscape nursery stock as described above. The automatic fertilization and watering control system for landscape nursery stock includes:

[0055] A soil state characteristic monitoring module, which is used to collect nursery stock morphological data for landscape nursery stock through electronic monitoring to obtain nursery stock morphological data; deploy soil monitoring sensors under each landscape nursery stock, and perform soil state characteristic monitoring through the soil detection sensors to obtain soil state characteristic monitoring feature data;

[0056] A root absorption limitation estimation module, which is used to predict the density of nursery stock roots based on the nursery stock morphological data to generate nursery stock root density prediction data; perform soil salt content abnormal fluctuation analysis on the soil state characteristic monitoring feature data to obtain soil salt content abnormal fluctuation data; perform root absorption limitation estimation on the nursery stock root density prediction data according to the soil salt content abnormal fluctuation data to obtain root absorption limitation estimation data;

[0057] The segmented drip irrigation ratio fusion module is used to perform slow-release fertilizer dosage control matching on the abnormal fluctuation data of soil salt content according to the root absorption limitation estimation data, and obtain the slow-release fertilizer dosage control data; based on the slow-release fertilizer dosage control data, perform segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio for fertilization and watering.

[0058] The fertilization and watering control model construction module is used to construct a fertilization and watering control model for the segmented drip irrigation ratio of fertilization and watering based on the policy gradient algorithm, and obtain the fertilization and watering control model for landscape nursery stock; send the fertilization and watering control model for landscape nursery stock to the terminal to execute the automatic fertilization and watering control of landscape nursery stock.

[0059] The beneficial effects of the present invention are as follows. By collecting the morphological data of landscape nursery stock and the characteristics of soil conditions through electronic monitoring and soil monitoring sensors, detailed information on the growth status of the nursery stock and the soil environment in which it is located can be accurately obtained. The morphological data of the nursery stock obtained through electronic monitoring provides a basis for predicting the root density in the subsequent stage, while the soil data provided by the soil monitoring sensors can help to grasp the important characteristics such as soil humidity, temperature, and salinity in real time, ensuring that the entire maintenance process is more accurate, avoiding errors caused by manual intervention, and improving the maintenance effect and resource utilization efficiency. Based on the morphological data of the nursery stock, the root density is predicted. By scientifically analyzing the morphological characteristics of the nursery stock, the distribution and density of the roots can be estimated, thereby providing a basis for evaluating the root absorption status. At the same time, after analyzing the abnormal fluctuations in the salt content of the soil condition characteristic monitoring data, the salt changes in the soil can be identified, helping to judge whether adverse conditions such as soil salinization occur. By combining these data to estimate the limited root absorption, the growth status of the nursery stock can be effectively grasped, providing a scientific basis for subsequent fertilization and watering, and avoiding the situation of excessive or insufficient fertilization. By matching the controlled release fertilizer dosage with the abnormal fluctuations in the soil salt content based on the estimated data of limited root absorption, the amount of fertilizer to be applied can be accurately calculated according to the salt content of the soil and the root absorption status, avoiding damage to the roots caused by excessive salt. The segmented drip irrigation ratio fusion technology can more reasonably manage the integration of water and fertilizer through precise control of the controlled release fertilizer dosage, ensuring that the nursery stock can grow healthily in a suitable environment, while avoiding waste of resources and improving the scientific nature and efficiency of garden maintenance. Based on the policy gradient algorithm, a fertilization and watering control model is constructed. Through the intelligent algorithm, the ratio of fertilization and watering can be automatically adjusted to ensure that the needs of the nursery stock at different growth stages are accurately met. The policy gradient algorithm continuously optimizes and adjusts the policy, improving the adaptability and decision-making accuracy of the system, and realizing dynamic management and real-time feedback. This automated control model can not only greatly improve the efficiency of garden fertilization and watering, but also achieve environmentally friendly goals such as water conservation and fertilizer conservation, and is executed through terminal devices to achieve precise and automated maintenance of landscape nursery stock. Therefore, the present invention is an optimized treatment of a traditional method for automatically controlling fertilization and watering of landscape nursery stock, solving the problem of inaccurate analysis of the limited root nutrient absorption caused by soil salinization in the traditional method for automatically controlling fertilization and watering of landscape nursery stock, resulting in large errors in automated fertilization and watering, improving the accuracy of analyzing the limited root nutrient absorption caused by soil salinization, and reducing the errors caused by automated fertilization and watering. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a schematic flow chart of the steps of a method for automatically controlling fertilization and watering of landscape nursery stock;

[0061] Figure 2 isFigure 1 Schematic diagram of the detailed implementation steps of step S2 in

[0062] Figure 3 is Figure 1 Schematic diagram of the detailed implementation steps of step S3 in Specific implementation manner

[0063] Please refer to Figures 1 to 3 , an automatic fertilization and watering control method for landscape nursery stock, the method comprising the following steps:

[0064] Step S1: Collect nursery stock morphological data of landscape nursery stock through electronic monitoring to obtain nursery stock morphological data; deploy soil monitoring sensors under each landscape nursery stock, and monitor the soil state characteristics through the soil detection sensors to obtain soil state characteristic monitoring feature data;

[0065] Step S2: Predict the density of nursery stock roots based on the nursery stock morphological data to generate nursery stock root density prediction data; analyze the abnormal fluctuations of soil salt content in the soil state characteristic monitoring feature data to obtain soil salt content abnormal fluctuation data; estimate the root absorption limitation according to the soil salt content abnormal fluctuation data for the nursery stock root density prediction data to obtain root absorption limitation estimation data;

[0066] Step S3: Control and match the slow-release fertilizer dosage according to the root absorption limitation estimation data for the soil salt content abnormal fluctuation data to obtain slow-release fertilizer dosage control data; perform segmented drip irrigation ratio fusion based on the slow-release fertilizer dosage control data to obtain the fertilization and watering segmented drip irrigation ratio;

[0067] Step S4: Construct a fertilization and watering control model for the fertilization and watering segmented drip irrigation ratio based on the policy gradient algorithm to obtain a landscape nursery stock fertilization and watering control model; send the landscape nursery stock fertilization and watering control model to the terminal to perform automatic fertilization and watering control of landscape nursery stock.

[0068] In the embodiment of the present invention, referring to Figure 1 as described, it is a schematic diagram of the step flow of an automatic fertilization and watering control method for landscape nursery stock of the present invention. In this example, the automatic fertilization and watering control method for landscape nursery stock comprises the following steps:

[0069] Step S1: Collect nursery stock morphological data of landscape nursery stock through electronic monitoring to obtain nursery stock morphological data; deploy soil monitoring sensors under each landscape nursery stock, and monitor the soil state characteristics through the soil detection sensors to obtain soil state characteristic monitoring feature data;

[0070] In the embodiments of the present invention, morphological data of landscape nursery stock is collected by a 4K high-definition camera installed 3 meters above the nursery area. The camera takes pictures at a frequency of once every 2 hours, with a shooting angle of 45° from above. The resolution is set to 3840×2160 pixels. The obtained images are processed by an image segmentation algorithm to extract three key morphological indicators, namely, the height of the nursery stock, the crown diameter, and the leaf color index, forming a morphological data matrix of the nursery stock. At the same time, a plurality of soil monitoring sensors are arranged in a regular pentagon around the root of each landscape nursery stock. The burial depth of the sensors is 15 cm, and the distance between each sensor is 25 cm. Each sensor collects data on soil pH value (accuracy ±0.1), water content (accuracy ±2%), temperature (accuracy ±0.3°C), conductivity (accuracy ±3%), and organic matter content (accuracy ±5%) every 30 minutes. The collected raw data is transmitted to the data processing terminal through a wireless transmission module using the ZigBee protocol. The data processing terminal first eliminates outliers in the collected data according to the 3σ principle, then linearly interpolates to fill in the missing data points, and then removes high-frequency noise through wavelet transform. Finally, the time-series fluctuation characteristics, spatial distribution characteristics, and correlation characteristics of the soil state characteristics are extracted, forming a soil state characteristics monitoring feature data set containing 15 feature dimensions, providing a data basis for subsequent root system prediction and soil analysis.

[0071] Step S2: Predict the root density of the nursery stock based on the morphological data of the nursery stock to generate root density prediction data of the nursery stock; analyze the abnormal fluctuations of the soil salt content in the soil state characteristic monitoring feature data to obtain abnormal fluctuation data of the soil salt content; estimate the root absorption limitation of the root density prediction data of the nursery stock according to the abnormal fluctuation data of the soil salt content to obtain root absorption limitation estimation data;

[0072] In the embodiments of the present invention, first, based on the seedling morphological data obtained in step S1, a morphology-root system correlation algorithm is applied to calculate the root system density of the seedlings. This algorithm uses the ratio of the crown diameter to the seedling height as the root system lateral expansion index (the value range is 0.5 - 2.5), and the leaf color index as the root system vitality coefficient (the value range is 0.6 - 1.2). The product of the two is then multiplied by the root system coefficient corresponding to the seedling type (1.0 for arbors, 1.2 for shrubs, 0.8 for vines) to obtain the predicted value of the root system density (the value range is 0.24 - 3.0). At the same time, time-frequency analysis is performed on the conductivity data in the monitoring characteristic data of the soil state characteristics. The main frequency and amplitude of the conductivity fluctuation are extracted through Fourier transform. When the fluctuation amplitude exceeds 25% of the reference value and the duration exceeds 72 hours, it is determined that there is an abnormal fluctuation in the soil salt content, and the abnormal fluctuation index is calculated (fluctuation amplitude percentage × duration / 24). Then, ion concentration decomposition processing is carried out, and the conductivity data is decomposed into ion concentration time series data according to the contribution rates of five main ions, namely Na+, Cl-, K+, Ca2+, and Mg2+ (0.35, 0.30, 0.15, 0.12, and 0.08 respectively). Then, the cell membrane damage intensity is calculated based on the ion concentration data, and the exponential regression equation D = 0.15×e^(0.28×S) is used (D is the damage intensity, S is the salinity value). At the same time, based on the analysis of the fluctuation data of the soil organic matter content, the nutrient leakage amount is analyzed, and the linear relationship model L = 0.2×D×O is used (L is the leakage amount, D is the damage intensity, O is the organic matter content). Considering comprehensively the predicted data of the root system density, the increment data of the ion competition intensity (obtained by the mutation detection algorithm), and the data of soil nutrient imbalance, the weighted summation method (weights are 0.4, 0.3, and 0.3 respectively) is used to obtain the estimated data of the root system absorption limitation, and the numerical range is 0 - 1, where 0 means no limitation and 1 means complete limitation.

[0073] Step S3: According to the estimated data of the root system absorption limitation, the abnormal fluctuation data of the soil salt content is controlled and matched for the slow-release fertilizer dosage to obtain the slow-release fertilizer dosage control data; based on the slow-release fertilizer dosage control data, segmented drip irrigation ratio fusion is carried out to obtain the segmented drip irrigation ratio of fertilization and watering;

[0074] In the embodiments of the present invention, based on the estimated data of limited root absorption and the abnormal fluctuation data of soil salt content obtained in step S2, a matching calculation for the controlled dosage of slow-release fertilizer is performed, specifically using a piecewise linear function: when the estimated value R of limited root absorption is less than 0.3, the dosage of slow-release fertilizer is the reference amount (0.5 kg / m²) × (1 - R / 0.3); when 0.3 ≤ R < 0.7, the dosage of slow-release fertilizer is the reference amount × 0.5 × (0.7 - R) / 0.4; when R ≥ 0.7, the dosage of slow-release fertilizer is 0; at the same time, in combination with the fluctuation index W in the abnormal fluctuation data of soil salt content, when W < 10, the dosage of slow-release fertilizer remains unchanged; when 10 ≤ W < 30, the dosage of slow-release fertilizer is reduced by W - 10%; when W ≥ 30, the dosage of slow-release fertilizer is reduced by 20%; the minimum value of the two is taken as the final controlled dosage data of slow-release fertilizer; then, a learning of the dosage control logic for the controlled dosage data of slow-release fertilizer is carried out, and a decision tree is constructed using historical data. The input features are the estimated value of limited root absorption, the abnormal fluctuation index of soil salt content, the season factor (spring 1.2, summer 1.0, autumn 0.8, winter 0.6), and the seedling type factor (tree 1.0, shrub 0.8, vine 1.2), and the output is the correction coefficient K of the slow-release fertilizer dosage (the value range is 0.7 - 1.3); then, based on the estimated data of limited root absorption, the respiratory conversion demand space T = 0.8 × (1 - R)^2 is calculated (T is the respiratory oxygen demand, and R is the estimated value of limitation), the nutrient absorption rate interval A per unit time is calculated as A = [0.05 × (1 - R), 0.15 × (1 - R)] (A is the absorption rate interval, and the unit is mg / h·g root), and at the same time, the fertilizer efficiency dosage release rate F = 0.025 × K is calculated (F is the release rate, and the unit is mg / h·g fertilizer); finally, the data of T, A, and F are interpolated through time series to obtain the values at 365 time points within a 24-hour cycle, and the drip irrigation fertilization concentration ratio B = F / A median at each time point is determined according to the ratio of A to F, and a segmented setting is carried out: the 24 hours are divided into 6 time periods, namely 0:00 - 4:00, 4:00 - 8:00, 8:00 - 12:00, 12:00 - 16:00, 16:00 - 20:00, 20:00 - 24:00, and the average value of the B value within each time period is taken as the drip irrigation ratio value of that time period, forming a segmented drip irrigation ratio scheme for fertilization and watering.

[0075] Step S4: Based on the policy gradient algorithm, a fertilization and watering control model for the segmented drip irrigation ratio of fertilization and watering is constructed to obtain a fertilization and watering control model for landscape nursery stock; the fertilization and watering control model for landscape nursery stock is sent to the terminal to perform automatic fertilization and watering control for landscape nursery stock.

[0076] In the embodiment of the present invention, first, the fertilization and watering segmented drip irrigation ratio obtained in step S3 is normalized. The Min-Max normalization method is used to map the drip irrigation ratio values in each period to the interval [0.1, 1]. The formula is B'=(B - Bmin) / (Bmax - Bmin)×0.9 + 0.1 (where B' is the normalized ratio value, B is the original ratio value, Bmin is the minimum ratio value, and Bmax is the maximum ratio value), to obtain the normalized ratio of fertilization and watering segmented drip irrigation. Then, a fertilization and watering control model is constructed based on the policy gradient algorithm. This algorithm first defines the state space as four dimensions: the current period index, soil humidity, soil conductivity, and the estimated value of root absorption limitation. The action space is two dimensions: the drip irrigation switch state (0 or 1) and the drip irrigation flow rate (ranging from 10% to 100% of the reference flow rate). The reward function is the matching degree between the root absorption amount and the release amount (the calculation formula is 1 - |A actual - F actual| / max(A actual, F actual)). Next, a two-layer neural network is constructed as a policy function approximator. The first layer contains 16 neurons, the activation function is ReLU, the second layer contains 8 neurons, the activation function is ReLU, and the output layer has 2 neurons corresponding to the two dimensions of the action space respectively. Through training in the simulation environment for 500 rounds, each round contains 24 periods, the learning rate is set to 0.01, and the discount factor is set to 0.95, to obtain the fertilization and watering control policy π(a|s) (indicating the probability of selecting action a in state s). The policy function is converted into a decision table form, including all discrete combinations of the state space (6 kinds of period indexes, 5 grades of soil humidity, 5 grades of soil conductivity, and 5 grades of the estimated value of root absorption limitation) and the corresponding optimal actions, totaling 750 decision rules, to form the fertilization and watering control model for landscape nursery stock. Finally, the control model is packaged into a binary file and transmitted to the on-site drip irrigation control terminal device through the 4G network. The control terminal executes the corresponding drip irrigation control instructions according to the real-time monitoring data by looking up the table, completing the automated fertilization and watering control of landscape nursery stock. The control terminal detects the environmental state every 10 minutes and updates the control decision to achieve all-weather precise fertilization and watering.

[0077] Step S1 includes the following steps:

[0078] Step S11: Collect the nursery stock morphological data of the landscape nursery stock through electronic monitoring to obtain the nursery stock morphological data;

[0079] Step S12: Deploy soil monitoring sensors under each landscape nursery stock, and monitor the soil state characteristics through the soil detection sensors to obtain the soil state characteristic monitoring data;

[0080] Step S13: Clean the soil state characteristic monitoring data to obtain the cleaned soil state characteristic monitoring data;

[0081] Step S14: Analyze the state characteristic features of the soil state characteristic monitoring cleaning data to obtain the soil state characteristic monitoring feature data.

[0082] In the embodiments of the present invention, the morphological data of nursery stock is collected through an electronic monitoring system installed in the planting area of landscape nursery stock. The electronic monitoring system consists of 8 high-definition digital cameras fixed on brackets at a height of 3.5 meters. The coverage area of each camera is 100 square meters. The resolution of the camera is 2560×1440 pixels, and the frame rate is 25 frames per second. The camera uses a 120° wide-angle lens and is equipped with an infrared supplementary lighting device to meet the all-weather monitoring requirements. The nursery stock is photographed at four fixed time points of 6:00, 10:00, 14:00, and 18:00 every day to obtain image data of 8 angles, including the top view and side view of the nursery stock. At the same time, the system is also equipped with a multi-spectral imaging device, which uses 5 bands (blue light at 450nm, green light at 550nm, red light at 650nm, near-infrared at 760nm, and short-wave infrared at 900nm) for imaging. The multi-spectral image is collected once a week, and the resolution of the multi-spectral image is 1280×960 pixels. All image data is transmitted to the data processing server in real time through the 5G network. After being processed by the image segmentation algorithm, morphological parameters such as the height of the nursery stock (accurate to 0.1 cm), the crown diameter (accurate to 0.5 cm), the thickness of the stem (accurate to 0.05 cm), the number of leaves (accurate to 10 pieces), the leaf color index (RGB value in the range of 0-255), and the crown volume (accurate to 0.01 cubic meters) are extracted to form a nursery stock morphological data matrix containing 6 dimensions and data at 4 time points. Each element in the data matrix is marked with a timestamp to track the changes in the growth status of the nursery stock.Deploy detection sensors and monitor data for the soil area beneath each landscape nursery stock. The specific approach is to arrange a soil monitoring sensor network within the root range of each nursery stock (usually 1.2 times the crown projection area). This network consists of a central main sensor and four sub-sensors, distributed in a "plus" shape. The central main sensor is located 10 cm directly below the main trunk of the nursery stock, and the four sub-sensors are respectively located 30 cm in the north, east, south, and west directions of the main trunk. The sensors are buried at two depths of 5 cm and 20 cm; the soil monitoring sensors used are multi-parameter integrated digital sensor modules. The sensor probe has a length of 25 cm and a diameter of 2 cm, and includes an integrated circuit module for measuring soil pH value (measurement range 3 - 9, accuracy ±0.1), soil water content (measurement range 5% - 60%, accuracy ±2%), soil temperature (measurement range -20°C to 50°C, accuracy ±0.5°C), soil conductivity (measurement range 0 - 5 mS / cm, accuracy ±3%), and soil organic matter content (measurement range 0 - 10%, accuracy ±5%); the sensors collect data once every 15 minutes according to a preset program, and the data is transmitted to the on-site data collector through the ZigBee wireless transmission protocol (2.4 GHz frequency band, transmission rate 250 kbps, transmission distance 100 m). The data collector uploads the data packets to the cloud server for storage through the 4G network; within 24 hours, a total of 96 time points and 10 spatial points of 5 parameter data are collected, forming a soil state characteristic monitoring data set of 4800 original data points for each nursery stock per day.

[0083] Perform comprehensive data cleaning on the soil status characteristic monitoring data obtained in step S12. First, conduct a data integrity check. Evaluate the data quality by detecting the data missing rate (the calculation method is the number of missing data points / the total number of data points × 100%). When the missing rate is lower than 5%, use the linear interpolation method to complete the missing data. The specific operation is to calculate the average change rate by taking the monitoring values of the two time points before and after the missing point, and then calculate the interpolation according to the time interval. When the missing rate is between 5% and 15%, use the polynomial interpolation method. Take the monitoring values of the 3 time points before and after the missing point to fit a cubic polynomial curve, and calculate the function value corresponding to the missing time point. When the missing rate exceeds 15%, mark the monitoring point as an invalid point. Then, conduct outlier detection. Adopt the 3σ criterion (that is, when the data deviates from the mean by more than 3 times the standard deviation, it is determined as an outlier). The calculation method is to first calculate the mean μ and standard deviation σ of the data for 24 consecutive hours at the same parameter and the same spatial point, and then detect whether each data point x satisfies |x - μ| > 3σ. For the detected outliers, use median filtering for replacement, and the replacement value is the median of the 2 time points before and after the outlier. Then, conduct data smoothing. Adopt the moving window average method, and set the window size to 5 time points (that is, 75 minutes). Take the arithmetic mean of the data within the window as the smoothed value of the center point. Finally, conduct data normalization. For different parameters, use the Min - Max normalization method to map the data to the [0, 1] interval. The calculation formula is x' = (x - xmin) / (xmax - xmin), where x is the original data, xmin and xmax are the minimum and maximum values of the parameter respectively, and x' is the normalized data value. After the above four - step processing, the obtained soil status characteristic monitoring cleaned data is more regular and coherent, laying a foundation for subsequent feature analysis.Perform multi-dimensional feature extraction and analysis on the soil state characteristic monitoring and cleaning data obtained in step S13. First, perform time series feature analysis to extract the time variation characteristics of each monitoring parameter, including the intraday fluctuation range (maximum value minus minimum value), intraday volatility (fluctuation range / mean value × 100%), peak occurrence time point (value range 0 - 23 hours), trough occurrence time point (value range 0 - 23 hours), and fluctuation periodicity (calculated by the autocorrelation function, value range 0 - 1) of the five time series features; then perform spatial distribution feature analysis to calculate the spatial distribution characteristics of the data at 10 spatial points, including the horizontal gradient (parameter difference between monitoring points in the east-west direction / distance, unit: parameter unit / meter), vertical gradient (parameter difference between monitoring points at different depths at the same location / depth difference, unit: parameter unit / meter), and spatial inhomogeneity (standard deviation / mean value × 100%) of the three spatial features; then perform parameter correlation analysis to calculate the correlation between different monitoring parameters through the Pearson correlation coefficient. The calculation method is the covariance of the two parameters divided by the product of the standard deviations of the two parameters. The obtained correlation coefficient matrix reflects the correlation strength between the five parameters of pH value, water content, temperature, conductivity, and organic matter content (value range -1 to 1); finally, integrate all the extracted features to form the soil state characteristic monitoring feature data for each seedling. This data includes time series features (5 parameters × 5 time series features = 25 feature dimensions), spatial distribution features (5 parameters × 3 spatial features = 15 feature dimensions), and parameter correlation features (10 pairs of pairwise combinations of 5 parameters × 1 correlation coefficient = 10 feature dimensions), a total of 50 feature dimensions of the feature vector. Each dimension in the feature vector is given a clear physical meaning, providing comprehensive and accurate feature data support for the subsequent prediction of the root density of seedlings and the analysis of abnormal fluctuations in soil salt content.

[0084] Step S2 includes the following steps:

[0085] Step S21: Predict the root density of seedlings based on the seedling morphology data to generate the predicted data of the root density of seedlings;

[0086] Step S22: Analyze the abnormal fluctuations in soil salt content of the soil state characteristic monitoring feature data to obtain the abnormal fluctuation data of soil salt content;

[0087] Step S23: Analyze the effectiveness loss of soil nutrient imbalance of the soil state characteristic monitoring feature data based on the abnormal fluctuation data of soil salt content to obtain the effectiveness loss data of soil nutrient imbalance;

[0088] Step S24: Estimate the root absorption limitation of the predicted data of the seedling root density based on the abnormal fluctuation data of soil salt content and the effective loss data of soil nutrient imbalance, so as to obtain the estimated data of root absorption limitation.

[0089] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0090] Step S21: Predict the seedling root density based on the seedling morphological data, and generate the predicted data of the seedling root density;

[0091] In the embodiment of the present invention, the seedling root density is predicted based on the seedling morphological data obtained in step S1. First, six parameters in the seedling morphological data, namely seedling height, crown diameter, stem thickness, number of leaves, leaf color index, and crown volume, are preprocessed. Each parameter is standardized into a standard normal distribution form with a mean of 0 and a standard deviation of 1 through the Z-Score standardization method. The calculation formula is z = (x - μ) / σ, where x is the original parameter value, μ is the mean of the parameter, and σ is the standard deviation of the parameter. Then, a root density prediction algorithm is constructed. This algorithm is based on the correlation between morphological parameters and root development and adopts a weighted summation calculation method. Specifically: the root density index , where D is the crown diameter (unit: centimeter), H is the seedling height (unit: centimeter), S is the stem thickness (unit: centimeter), L is the number of leaves (unit: piece), V is the crown volume (unit: cubic centimeter), C is the leaf color index (dimensionless value from 0 to 100, converted from RGB), to are weight coefficients, which are respectively 0.35, 0.25, 0.15, 0.15, and 0.10; for different types of seedlings, a species adjustment coefficient k is introduced. For coniferous seedlings, k = 0.8, for broad-leaved seedlings, k = 1.0, and for mixed seedlings, k = 0.9. The final predicted value of the root density is k × RDI; according to the calculated predicted value of the root density, it is divided into five levels: extremely low (0 - 0.2), low (0.2 - 0.4), medium (0.4 - 0.6), high (0.6 - 0.8), and extremely high (0.8 - 1.0). At the same time, a root distribution prediction map is generated, showing the root distribution density in the horizontal and vertical directions, and the density is represented by the depth of color, forming the predicted data of the seedling root density including numerical values and visualization results.

[0092] Step S22: Analyze the abnormal fluctuation of soil salt content for the monitoring characteristic data of soil state characteristics to obtain the abnormal fluctuation data of soil salt content;

[0093] In the embodiments of the present invention, for the analysis of abnormal fluctuations in soil salt content of the monitoring characteristic data of soil state characteristics, first, parameters related to soil salt content are extracted from the monitoring characteristic data of soil state characteristics, mainly including three indicators: soil conductivity, pH value, and organic matter content; then, time series analysis is performed on the soil conductivity data. Using the sliding window method, the window size is set to 72 hours (i.e., 288 consecutive data points), and the window sliding step is 6 hours (i.e., 24 data points). Calculate the mean value, standard deviation, change rate (the difference between the first and last values / mean value × 100%), and fluctuation frequency (the number of peak occurrences in the window / 3) of the conductivity within each window; then introduce the abnormal salt content determination criterion. When any of the following conditions is met, it is determined as an abnormal fluctuation: ① The conductivity continuously rises within 72 hours and the change rate exceeds 25%; ② The conductivity suddenly increases by more than 30% of the reference value (the average value of the previous 7 days) within 24 hours; ③ The conductivity fluctuation frequency is abnormal (the number of fluctuations within 24 hours exceeds 6 times and the amplitude of each fluctuation exceeds 10%); ④ The conductivity and pH value show a reverse and drastic change (the conductivity rises while the pH value drops, and the change amplitude of both exceeds 20%); then, quantify the detected abnormal fluctuations and calculate the abnormal fluctuation index , where ΔEC is the conductivity change rate, T is the abnormal duration (in hours), F is the fluctuation frequency, ΔpH is the pH value change rate, to are weight coefficients, with values of 0.4, 0.3, 0.2, and 0.1 respectively; finally, according to the abnormal fluctuation index AWI, the abnormal fluctuations are divided into four levels: mild (AWI < 30), moderate (30 ≤ AWI < 60), severe (60 ≤ AWI < 90), and extremely severe (AWI ≥ 90), generating soil salt content abnormal fluctuation data including the abnormal fluctuation time period, abnormal fluctuation index, fluctuation level, and fluctuation trend graph, providing a basis for subsequent estimation of restricted root absorption.

[0094] Step S23: Based on the soil salt content abnormal fluctuation data, perform an analysis of the effectiveness loss of soil nutrient imbalance on the monitoring characteristic data of soil state characteristics to obtain soil nutrient imbalance effectiveness loss data;

[0095] In the embodiments of the present invention, for the analysis of the effectiveness loss of soil nutrient imbalance on the monitoring characteristic data of soil state characteristics based on the soil salt content abnormal fluctuation data, first, convert the conductivity data in the soil salt content abnormal fluctuation data into salt concentration data. The conversion formula is S = k × EC, where S is the salt concentration (in g / L), EC is the conductivity (in mS / cm), and k is the conversion coefficient, with a value of 0.64; then, decompose the salt concentration data into ion compositions. According to the typical soil salt composition ratio, decompose the total salt concentration into (accounting for 35%), (accounting for 27%), (accounting for 15%), (accounting for 12%), (accounting for 8%) and (accounting for 3%) of the six main ions to form the time - series change data of the salt - containing ion concentration; then, perform fluctuation extraction on the organic matter content data in the monitoring characteristic data of the soil state characteristics, and calculate the change amount of the organic matter content within 72 consecutive hours , where is the organic matter content at the starting moment, is the organic matter content after 72 hours. When ΔOM < 0, it indicates the degradation or loss of organic matter. Extract the data of these degradation or loss periods to form the soil organic matter content fluctuation data; then, establish a non - linear regression model for cell membrane damage based on the time - series change data of the salt - containing ion concentration, in the form of an exponential function , where MD is the degree of membrane damage (a dimensionless value between 0 and 1), S is the total salt concentration, and a and b are regression coefficients, taking values of 0.12 and 1.35 respectively. Calculate the degree of cell membrane damage under different salt concentrations through this model to form the cell membrane damage intensity regression data; then, combine the cell membrane damage intensity regression data with the soil organic matter content fluctuation data, and calculate the nutrient leakage amount NL = MD×OM×c, where NL is the nutrient leakage amount (unit: mg / kg), MD is the degree of membrane damage, OM is the organic matter content (unit: g / kg), and c is the conversion coefficient, taking a value of 8.5. Obtain the nutrient leakage equivalent prediction data through this calculation; finally, comprehensively analyze the results of various analyses and calculate the soil nutrient imbalance effectiveness loss index , where NL is the nutrient leakage amount, IC is the ion competition intensity increment data (obtained from step S235), pH_dev is the degree of deviation of the pH value from the optimal range (|pH - 6.5| / 6.5×100%), NP is the total soil nutrient amount (unit: mg / kg), to are the weight coefficients, taking values of 0.5, 0.3, and 0.2 respectively. The finally obtained NIL value range is 0 - 100%, indicating the percentage of soil nutrient effectiveness loss, and forming the soil nutrient imbalance effectiveness loss data.

[0096] Step S24: Estimate the root absorption limitation for the predicted data of the seedling root density according to the abnormal fluctuation data of the soil salt content and the soil nutrient imbalance effectiveness loss data to obtain the root absorption limitation estimation data.

[0097] In the embodiment of the present invention, the root absorption limitation of the predicted data of the seedling root density is estimated according to the abnormal fluctuation data of the soil salt content and the effective loss data of the soil nutrient imbalance. First, the fluctuation amplitude deviation rate operation is performed on the abnormal fluctuation data of the soil salt content. The calculation formula is SDF = (EC_max - EC_base) / EC_base × 100%, where SDF is the difference in the salt content fluctuation amplitude (unit: %), EC_max is the highest conductivity value during the monitoring period, and EC_base is the reference conductivity value (the average value of the previous 15 days); then, the increase ratio of the soil solution osmotic pressure is calculated according to the difference data of the salt content fluctuation amplitude. The van't Hoff equation π = i×C×R×T is used, where π is the osmotic pressure (unit: MPa), i is the degree of dissociation (the average value of salts is taken as 1.8), C is the salt concentration (unit: mol / L, obtained by converting the conductivity, and the conversion coefficient is 0.011), R is the gas constant (0.00831 L·MPa / (mol·K)), and T is the absolute temperature (taken as 293K). The increase ratio of the osmotic pressure OPI = (π_abnormal - π_normal) / π_normal × 100% is calculated, where π_abnormal is the osmotic pressure in the abnormal state and π_normal is the osmotic pressure in the normal state; then, the inhibition evaluation of the microbial nutrient conversion is carried out based on the increase ratio of the soil solution osmotic pressure. The microbial activity inhibition rate , where is the proportionality coefficient, with a value of 0.022, is the exponential coefficient, with a value of 1.25. The cumulative inhibition effect during the entire monitoring period is calculated by numerical integration ∫MIR(t)dt to obtain the nutrient conversion inhibition fluctuation value NTI; at the same time, the change of the soil structure is estimated according to the difference data of the salt content fluctuation amplitude and the increase ratio of the soil solution osmotic pressure. The soil compaction cumulative gradient SSG = α×SDF + β×OPI + γ×(SDF×OPI), where α, β, and γ are coefficients, with values of 0.015, 0.025, and 0.0005 respectively, to calculate the soil compaction cumulative gradient data; then, the restriction evaluation of the root growth and respiration is carried out based on the soil compaction cumulative gradient data. The restriction gradient calculation formula is RIG = SSG×(1 + δ×RDI), where RDI is the root density (obtained from step S21), and δ is the adjustment coefficient, with a value of 0.8, to obtain the root growth / respiration restriction gradient data; finally, considering the nutrient conversion inhibition fluctuation value, the root growth / respiration restriction gradient, and the effective loss data of the soil nutrient imbalance, the root absorption limitation coefficient is calculated , where is the weight coefficient, with values of 0.02, 0.03, and 0.025 respectively. The RAL value ranges from 0 to 1, where 0 indicates no restriction and 1 indicates complete restriction. By normalizing the RAL, estimation data on root absorption restriction with five levels (slight, mild, moderate, severe, and extremely severe) is finally generated.

[0098] Step S23 includes the following steps:

[0099] Step S231: Perform ion concentration decomposition on the abnormal fluctuation data of soil salt content to generate time-series change data of salt ion concentration;

[0100] Step S232: Extract the fluctuation of organic matter content from the monitoring characteristic data of soil state characteristics to obtain the fluctuation data of soil organic matter content;

[0101] Step S233: Conduct non-linear strength regression analysis of cell membrane damage based on the time-series change data of salt ion concentration to obtain the regression data of cell membrane damage strength;

[0102] Step S234: Perform an equivalent simulation prediction of organic matter nutrient leakage on the fluctuation data of soil organic matter content according to the regression data of cell membrane damage strength to obtain the equivalent prediction data of nutrient leakage;

[0103] Step S235: Analyze the increment of ion competition intensity on the fluctuation data of soil organic matter content according to the time-series change data of salt ion concentration to generate the increment data of ion competition intensity;

[0104] Step S236: Conduct an analysis of the effective loss of soil nutrient imbalance based on the equivalent prediction data of nutrient leakage and the increment data of ion competition intensity to obtain the data of the effective loss of soil nutrient imbalance.

[0105] In the embodiments of the present invention, the acquisition of abnormal fluctuation data of soil salt content is based on the conductivity time series monitoring curve. This curve collects 720 sample points within 120 hours at an interval of 10 minutes, and the unit is millisiemens per centimeter. When this data is processed by ion concentration decomposition, by setting common soil ion types, including five main ion components such as sodium ions, potassium ions, chloride ions, calcium ions, and magnesium ions, and applying the ion activity factor estimation method based on the extended Debye-Hückel equation, the total conductivity value is converted into the change values of various ion concentrations. This process uses a data window with a sliding window length of 12, and takes the conductivity fluctuation amplitude between the minimum value and the maximum value within each window as the standard deviation of the concentration change distribution, and combines local weighted regression (LOWESS) for ion concentration fitting. The output format is the concentration sequence of each ion at every 10-minute time point, with the unit of milligrams per liter, constituting the time series change data of salt-containing ion concentration. The monitoring characteristic data of soil state characteristics is used to extract the fluctuation of organic matter content. First, relying on the installed soil sensors, by monitoring the basic state characteristics of the soil such as water content, pH value, oxidation-reduction potential (ORP), etc., data related to soil health and nutrient activity is obtained. On this basis, for the fluctuation of the content of soil organic matter, data preprocessing is first carried out to eliminate noise data. Then, through dynamic time series analysis, the characteristic data of each period is smoothed to remove the interference of seasonal change factors and highlight the short-term volatility of organic matter in the soil. The wavelet transform method is used to perform multi-scale analysis on the soil data to capture the high-frequency components of the fluctuation of soil organic matter content. By comparing the change trends of organic matter in each period and combining the fluctuation characteristics of the data, the fluctuation data of soil organic matter content is generated. Within each time window, the output of the organic matter fluctuation data includes key characteristics such as fluctuation amplitude, fluctuation period, and fluctuation speed. All data is stored in the form of a time series for subsequent analysis and decision-making. For example, during the monitoring period from March 21st to March 23rd, the ORP value of the soil increased from -120 mV to -70 mV, and there were fluctuations in the water content, indicating that the decomposition and conversion rates of organic matter had changed. This fluctuation data can be used for subsequent soil improvement and fertilization plan adjustment.

[0106] Based on the time series change data of salt-containing ion concentration, a non-linear strength regression analysis of cell membrane damage is carried out. First, a relationship model between ion concentration and cell membrane damage is constructed in the form of an exponential function , where MD is the membrane damage intensity (dimensionless value, range 0-1), C_Na is concentration (unit mmol / L), C_Cl is concentration (unit mmol / L), and α, β, γ are non-linear regression coefficients, with values of 0.95, 0.068, and 0.072 respectively; then for each time point and Substitute the concentration data into the model to calculate the corresponding membrane damage intensity values, forming a time-series dataset of membrane damage intensity; then analyze the interaction between membrane damage intensity and other ion concentrations, and introduce an ion antagonism factor , where δ is the antagonism coefficient with a value of 0.15, and C_K, C_Ca, and C_Mg are respectively concentrations, and correct the membrane damage intensity ; then calculate the cumulative effect of membrane damage intensity using the weighted moving summation method , where i represents the time interval (in hours), w_i is the time weight coefficient, satisfying Σw_i = 1, and w_i decreases as i increases. The specific values are w_0 = 0.3, w_1 = 0.25, w_2 = 0.2, w_3 = 0.15, w_4 = 0.1; subsequently, divide the cumulative membrane damage intensity data into intervals, and divide the membrane damage intensity into five levels: slight (0 - 0.2), mild (0.2 - 0.4), moderate (0.4 - 0.6), severe (0.6 - 0.8), and extremely severe (0.8 - 1.0); finally, according to the interval division results, calculate the proportion of each level within each time period, generate a time-series diagram of the membrane damage intensity level distribution, and calculate the change trend of the membrane damage intensity through a sliding window (window size is 24 hours), forming a cell membrane damage intensity regression data including original data, level distribution, and trend analysis. In the operation process of performing an equivalent simulation prediction of organic matter nutrient leakage on the soil organic matter content fluctuation data based on the cell membrane damage intensity regression data, first, match the cell membrane damage intensity regression data with the soil organic matter concentration fluctuation data one by one, and establish an equivalent derivation relationship of leakage intensity using the differential and differential recurrence algorithm. In this operation process, set the time span to one cycle every 60 minutes, collect the mass concentration data of dissolved organic carbon (DOC) in the soil, and link it with the cell membrane damage intensity data. By performing time-series differential processing on the DOC data, extract the DOC change rate per unit area between two adjacent cycles, and perform leakage inference according to the cell membrane damage regression coefficient. The leakage inference algorithm used is the quantitative piecewise extrapolation method. For example, in a certain test area, the initial concentration of DOC in the surface soil is 65 mg / L, and the membrane damage intensity is 0.78. After one cycle, the concentration drops to 54 mg / L, and the extrapolation logical judgment is an effective leakage rather than the contribution of the mineralization process. This method is based on a fixed time difference, and couples the DOC concentration change and the cell membrane strength regression data in different cycle intervals according to the discrete integral method, and finally outputs the total DOC leakage per unit area per unit time as the equivalent prediction data of organic matter nutrient leakage, with the unit of mg / (m²·h). When performing an ion competition intensity increment analysis on the soil organic matter content fluctuation data based on the time-series change data of salt ion concentrations, first, based on each type of main cation (including , , , ) concentration change curve to establish a time series data matrix, and construct an analytic function for the incremental ion mutual repulsion rate. This function is based on the charge pair effect and the soil cation exchange capacity rule. The concentration change rate is the first driving factor, The concentration is the response factor, and the Beta distribution dependency regression method is used to derive the competition increment intensity. Among them, the data preprocessing method is standardized time series difference and peak calibration to eliminate the instantaneous fluctuation interference caused by environmental water input. In the specific experiment, the initial soil The concentration was 1.8 mmol / L, which increased to 3.1 mmol / L after 48 hours. The concentration dropped from 2.4 to 1.6 mmol / L, and the corresponding ion competition intensity increment value in the regression coupling curve was 0.58, which represents the adsorption priority transfer intensity between sodium and potassium. This process does not involve manual setting of thresholds, but is completely analyzed based on the time trend of ion concentration and the mutually exclusive comparison logic, and finally outputs the ion competition intensity increment data, in units of dimensionless relative gain ratio. In the operation of analyzing the effectiveness loss of soil nutrient imbalance based on nutrient leakage equivalent prediction data and ion competition intensity increment data, the leakage equivalent data of the three main nutrient elements (nitrogen, phosphorus, and potassium) are first used as the basic indicator items, corresponding to the ion competition intensity increment data of the same time period, and the imbalance loss function is constructed by the weighted integration method. The weight distribution logic adopts the standard three-factor evaluation model, in which the leakage equivalent weight is set to 0.6, and the competition intensity increment is set to 0.4. After integrating the data, the total effectiveness loss per unit area is obtained by solving the weighted difference within the time period. In this operation process, the data resolution is controlled to one grid area per 10 square meters, and one evaluation cycle is every 24 hours. The analysis indicators include nitrogen transfer ratio (according to concentration change), phosphorus release ratio (based on concentration and hydrolysis rate), potassium effective concentration inhibition ratio (based on and Relative change derivation). Taking the real collected data as an example, in a certain sampling, the nitrogen leakage was 13 mg / m2 per hour, and the ion competition increment value was 0.52. After the weighted function processing, the nitrogen effectiveness loss in this period was 10.14 mg / m2. This value was written into the database as the final generated soil nutrient imbalance effectiveness loss data and used in the subsequent irrigation and fertilization strategy optimization process. The entire operation process did not use neural networks or machine learning models, but was completed using static rule algorithms and mathematical derivation based on data distribution logic.

[0107] Step S235 includes the following steps:

[0108] Perform a mutation increment index calculation on the time-series change data of the salt ion concentration to obtain the salt ion concentration mutation increment index;

[0109] Based on the salt ion concentration mutation increment index, perform a continuous concentration mutation approximation combination on the time-series change data of the salt ion concentration to generate concentration continuous mutation approximation combination data;

[0110] Based on the concentration continuous mutation approximation combination data, perform a pre-emptive continuous intensity simulation of the salt ion absorption sites on the soil organic matter content fluctuation data to obtain the pre-emptive continuous intensity data of the salt ion absorption sites;

[0111] Perform an ion flux time-series interval difference calculation on the pre-emptive continuous intensity data of the absorption sites to obtain the salt ion flux difference of the absorption sites;

[0112] Perform an ion competition intensity increment analysis based on the salt ion flux difference of the absorption sites to generate ion competition intensity increment data.

[0113] In the embodiment of the present invention, for the operation of "performing a mutation increment index calculation on the time-series change data of the salt ion concentration to obtain the salt ion concentration mutation increment index", the acquisition path of the salt ion concentration needs to be clarified first. In this embodiment, a soil deep-layer multi-point ion-selective electrode sensor is used to collect the time-series change data sequence of the ion concentration. Before processing the original sequence, first perform a 5th-order sliding window denoising process on each ion concentration curve with Savitzky-Golay smoothing filtering, and retain the first derivative information for subsequent analysis. Subsequently, use the time-series difference increment method to sequentially obtain the change amplitude of the ion concentration at adjacent times, and construct an instantaneous mutation rate sequence of the ion concentration. To measure the non-linear degree of the mutation intensity, an exponential calculation is performed by accumulating the deviation between the amplitude ratio and the historical change mean. In the specific process, set the historical window length to 12 hours, accumulate the ratio difference between the current change value and the change mean within this window, and iteratively update hour by hour to form a mutation increment index curve. In actual operation, when the soil When the concentration rapidly increases from 23.4 mmol / L to 41.2 mmol / L between the 48th hour and the 51st hour and maintains at a high level, when the cumulative value of the mutation increment index within this window exceeds the threshold standard of 5.6, it is marked as a first-level concentration mutation node and incorporated into the subsequent approximation combination processing flow. For the operation of "conducting continuous concentration mutation approximation combination on the time-series change data of salt-containing ion concentration based on the mutation increment index of salt-containing ion concentration to generate concentration continuous mutation approximation combination data", first, the time interval marked as a first-level mutation node in the mutation increment index calculation is calibrated. Subsequently, a curve continuity reconstruction method based on B-spline interpolation is used to perform a fitting approximation operation on the ion concentration curve in the mutation section. To ensure that the fitted mutation form truly reflects the rapid change process, a third-order B-spline interpolation function is used, and boundary constraints on the mutation rate are set for the control nodes to ensure that there is an obvious discontinuous jump in the first derivative of the function at the nodes, and this jump amount is consistent with the original mutation increment. During actual processing, for the mutation section of the concentration from the 48th to the 51st hour, a fitting sequence with a fitting point every 30 minutes is generated through interpolation to form a concentration mutation combination data set containing the original concentration points and the fitting complement points. This combined data not only retains the original mutation information but also provides compensation for the change trend at non-sampled time points for use in the next absorption site intensity simulation. For the operation of "conducting a preemptive continuous intensity simulation of salt-containing ion absorption sites on the soil organic matter content fluctuation data based on the concentration continuous mutation approximation combination data to obtain the preemptive continuous intensity data of salt-containing ion absorption sites", the change data of the soil organic matter content in the same sampling time and space region needs to be introduced synchronously. The near-infrared reflectance spectroscopy method and the soil high-frequency mixing sampling method are used to obtain the organic matter content value once per hour. This data is aligned with the concentration mutation combination data generated above according to the timestamp, and a comparison relationship map of ion concentration and organic matter fluctuation is established under the same time axis. During the operation, the salt-containing ion preemption model is set as an interval superposition model constructed based on the order of concentration fluctuations. That is, within the mutation region, if the time when the ion concentration increases is earlier than the starting point of the decrease in the organic matter content, it is marked as the occurrence of a potential absorption site preemption phenomenon. For each comparison interval, the mutation duration and the concentration increase rate are set as weighting factors, and an intensity fitting function is constructed by linear superposition. In the above-mentioned 48th to 51st hour In the mutated segment, the organic matter content decreased from the original 2.3% to 1.7%, and the starting point of the decrease lagged by approximately 0.5 hours. This region constitutes a high-intensity preemptive absorption event, and the calculated intensity fitting value reaches the upper limit set by the preemption warning model. Finally, the continuous intensity simulation results form a time-series image with the time axis as the horizontal axis and the absorption intensity per unit time as the vertical axis, which is used as the trigger basis for the organic matter compensation strategy in the subsequent control system callback fertilization plan. No model prediction was introduced during the whole process, and only actual observation data interpolation reconstruction and causal time series superposition judgment were used. All processing algorithms were completed based on known mathematical analysis methods and signal processing functions, without involving non-physical basic modeling operations.

[0114] For the operation of "calculating the time interval difference of ion fluxes for the preemptive continuous intensity data of absorption sites to obtain the difference in salt ion fluxes at absorption sites", it is necessary to calculate the difference in flux changes at fixed time intervals based on the preemptive continuous intensity data of absorption sites obtained in the previous processing step. The core logic of the flux difference processing is to estimate the difference in the number of ions absorbed per unit area at the same absorption site at two consecutive time points within a fixed time step. In this embodiment, the absorption site is defined as a soil structure unit with significant ion exchange activity within a 1 cm × 1 cm area in the rhizosphere, and the flux per unit time is based on millimoles per square centimeter per hour as the basic unit. The time step is set to 1 hour, and the continuous intensity data comes from the simulated absorption intensity time series values of three main ions such as . Taking

[0115] For the operation of "analyzing the increment of ion competition intensity based on the difference in salt ion flux at the absorption site and generating the data of ion competition intensity increment", it is necessary to extract the flux difference vector of each ion between the same absorption sites at the same time node from the flux difference matrix, and deduce the competition relationship from the relative change trend between the vectors. In this embodiment, based on the fixed absorption site, a multi-ion flux difference ratio map is constructed hour by hour to determine whether there is an absorption competition phenomenon within a certain period. The criterion for judging the competition relationship is as follows: if the flux difference of a certain ion increases significantly, and the flux difference of another ion shows a synchronous decrease, and the change in their ratio exceeds the static competition threshold of 0.8, it is considered that there is a competitive absorption behavior. At the 49th hour, the flux difference is +1.5, the flux difference is -0.4, the flux difference is -0.6, and the change directions of the three are inconsistent, and and the absolute value of the flux difference ratio is 2.5, exceeding the static competition threshold, and it is determined as there is an absorption inhibition competition effect on

[0116] Step S24 includes the following steps:

[0117] Step S241: Perform a fluctuation amplitude deviation rate operation on the abnormal fluctuation data of soil salt content to obtain the salt content fluctuation amplitude difference data;

[0118] Step S242: Quantify the conversion of the increase ratio of soil solution osmotic pressure according to the salt content fluctuation amplitude difference data to obtain the increase ratio of soil solution osmotic pressure;

[0119] Step S243: Integrate the fluctuation of the microbial nutrient conversion inhibition value based on the increase ratio of soil solution osmotic pressure to obtain the nutrient conversion inhibition fluctuation value;

[0120] Step S244: Estimate the soil compaction accumulation gradient according to the salt content fluctuation amplitude difference data and the increase ratio of soil solution osmotic pressure to obtain the soil compaction accumulation gradient data;

[0121] Step S245: Based on the soil compaction cumulative gradient data, conduct a root growth / respiration restriction gradient assessment on the predicted data of the seedling root density to obtain the root growth / respiration restriction gradient data;

[0122] Step S246: According to the nutrient conversion inhibition fluctuation value, the root growth / respiration restriction gradient data, and the soil nutrient imbalance effectiveness loss data, conduct a root absorption limitation estimation on the predicted data of the seedling root density to obtain the root absorption limitation estimation data.

[0123] In the embodiment of the present invention, for the operation of step S241 "calculating the fluctuation amplitude offset rate of the abnormal fluctuation data of the soil salt content to obtain the fluctuation amplitude difference data of the salt content", it is necessary to extract the intensity difference caused by the abnormal fluctuation by calculating the fluctuation amplitude offset rate within the reference time window based on the soil profile salt content data sequence collected in multiple consecutive time periods. The sampling depth is set to the 0-20 cm area of ​​the soil surface, the sampling frequency is once every 2 hours, the observation period is 72 hours, and the conductivity sensor is used to record the conductivity value of the unit volume of soil solution in real time, with Siemens per centimeter as the unit. In the implementation process, the conductivity value of each sampling point is first converted into an approximate equivalent salt content value (grams per kilogram), and the conversion factor used is based on the previous field calibration experimental data. Then, a sliding window with a unit of 12 hours is constructed for the continuous time series, and the fluctuation amplitude of the difference between the maximum and minimum values ​​in each window is estimated, and then the fluctuation amplitude difference between adjacent windows is compared, and the fluctuation amplitude of the initial window is used as a reference standard to calculate the offset rate. The calculation formula of the offset rate is: the fluctuation amplitude of the latter window minus the fluctuation amplitude of the previous window divided by the fluctuation amplitude of the previous window. For example, in the observation area where the seedling number is M-23, the fluctuation amplitude of the first time window is 1.6 grams per kilogram, and the fluctuation amplitude of the second window is 2.3 grams per kilogram, then the offset rate is 0.4375. The offset rate data is used to determine whether there is a nonlinear aggravated salt disturbance phenomenon in the area. The offset rate results of all observation points are finally summarized to generate a two-dimensional grid structure of salt content fluctuation amplitude difference data map, whose rows and columns correspond to the sampling point number and time index number. For the operation of step S242 "converting and quantifying the increase ratio of soil solution osmotic pressure according to the salt content fluctuation amplitude difference data to obtain the increase ratio of soil solution osmotic pressure", it is necessary to use the quantitative relationship between salt concentration and osmotic pressure, and complete the conversion based on the physical relationship that osmotic pressure and total solute concentration are approximately linearly positively correlated without introducing an empirical model. In the specific operation, first, from the salt content fluctuation difference data extracted in step S241, the time period and sampling point combination with an offset rate greater than 0.25 are screened, and then the corresponding salt change is substituted into the osmotic pressure conversion factor for direct conversion. When sodium chloride is the main salt ion, the osmotic pressure conversion factor is 0.036 MPa per gram per kilogram, which is obtained by repeatedly measuring the artificial proportion of NaCl solution in a standard constant temperature chamber. In the same area where the seedlings are numbered M-23, if the salt increases by 0.9 grams per kilogram in a certain time period, the increase in the osmotic pressure of the soil solution at this point is 0.0324 MPa. The osmotic pressure increment value is divided by the initial average osmotic pressure at this point (set to 0.142 MPa), and the osmotic pressure increase ratio can be obtained, and the result is 0.228.This operation is performed in parallel at all points where the offset rate exceeds the standard, generating a dataset of elevation ratio tensors with sampling points and time periods as dimensions. This data is used to describe the trend of potential osmotic pressure changes in the soil caused by salt fluctuations. All calculation processes rely on the quantitative correspondence between measured conductivity, salt content, and osmotic pressure, avoiding structural errors introduced by undefined empirical models. For the operation of step S243, "Integrate the numerical fluctuations of microbial nutrient transformation inhibition based on the elevation ratio of soil solution osmotic pressure to obtain the fluctuation value of nutrient transformation inhibition", based on the experimental evidence of the sensitivity of soil microorganisms to high osmotic pressure environments, a response inhibition function based on a weight factor is used to map and cumulatively integrate the elevation ratio of osmotic pressure to construct a dynamic weakening trend of nutrient transformation rate. In the operation, first, the main soil functional bacteria represented by nitrifying bacteria, phosphomonoesterase bacteria, and cellulose-degrading bacteria are set, and the proportion of metabolic activity decline at different osmotic pressure levels is measured under experimental conditions. It is known that when the elevation ratio of osmotic pressure is 0.2, the nitrification rate of nitrifying bacteria decreases by 22%, and the phosphatase enzyme activity decreases by 17%. In this embodiment, this relationship is used to assign weights to transformation inhibition. Subsequently, on an hourly basis, according to the elevation ratio tensor of soil solution osmotic pressure, numerical integration of the inhibition response of each corresponding bacterial group in each time period is performed. The integration operation takes time as the horizontal axis and the inhibition intensity as the vertical axis, and the integration result is the fluctuation value of nutrient transformation inhibition. Taking the data within the 72-hour cycle at point M-23 as an example, the overall integration value is 6.72, indicating that the cumulative average inhibition intensity caused by the increase in osmotic pressure throughout the cycle reaches 6.72 units. This result is used for the fine-tuning strategy of the target nutrient release rate in subsequent fertilization instructions to ensure the stability of the control system response under the condition of limited bacterial group functions. Throughout the process, the inhibition factors of each bacterial group are determined through field cultivation and controlled variable experiments, without introducing predictive inferences or hypothetical models. All calculation links are clear numerical logic and integral accumulation processes.

[0124] For the operation of step S244 "estimate the soil compaction cumulative gradient according to the salt content fluctuation amplitude difference data and the soil solution osmotic pressure increase ratio to obtain the soil compaction cumulative gradient data", based on two core input variables, namely the salt content fluctuation amplitude difference data obtained in step S241 and the osmotic pressure increase ratio tensor generated in step S242, a synchronous pairing relationship is established with the spatial position index and the time series index respectively, and the influence trend of the salt fluctuation amplitude and osmotic pressure change on the soil structure is synchronously analyzed in the same time period and sampling point. During the operation, the historical salt fluctuation value sequence of each grid is positively accumulated with a grid area of ​​0.5 meters by 0.5 meters as a unit to generate the fluctuation intensity accumulation; at the same time, the osmotic pressure increase ratio multiplied by the time span is used as the unit inhibition index, which is superimposed on the grid node to represent the induced trend of long-term osmotic pressure changes on the capillary water migration and particle structure convergence process. Subsequently, the salt fluctuation accumulation and the osmotic pressure inhibition index are standardized and linearly weighted, with the weights of 0.6 and 0.4 respectively, and the result is the initial estimation of the soil compaction cumulative gradient. Taking the garden nursery monitoring area numbered Z-11 as an example, the cumulative value of the salt content fluctuation at this point in a 7-day cycle is 7.2 grams per kilogram, the average osmotic pressure increase ratio is 0.27, and the inhibition index obtained by combining the time span of 168 hours is 45.36. The two values ​​after standardization are 0.81 and 0.63 respectively, and the hardening gradient obtained after linear weighting is 0.732. This operation is performed synchronously at all monitoring points, and finally a two-dimensional space-time hardening gradient tensor data set is formed, in which each unit contains the hardening trend data driven by salt accumulation and osmotic pressure, which is directly used as the input basis for subsequent root growth restriction analysis. For the operation of step S245 "based on the soil hardening cumulative gradient data, the root density prediction data of the seedlings is evaluated for the root growth / respiration restriction gradient, and the root growth / respiration restriction gradient data is obtained", it is necessary to first perform structured mapping on the seedling root density prediction data so that it maintains a spatial correspondence with the hardening cumulative gradient tensor obtained in step S244. The root density prediction data is the root volume data per unit volume (in grams per cubic decimeter) generated by the root shadow distribution range and underground biomass conversion ratio previously fused through image features. The spatial resolution is 0.5 meters, and the depth levels are divided from 0-10 cm, 10-20 cm, and 20-30 cm. In implementation, the root density value of each spatial unit is cross-analyzed with the cumulative gradient of the hardening of the unit where it is located, and the root growth capacity is converted by setting the growth inhibition coefficient. The inhibition coefficient is segmented based on the root density threshold and the hardening gradient threshold. If the root density is higher than 3.5 grams per cubic decimeter and the hardening gradient exceeds 0.7, the point is given a growth inhibition intensity of 0.6, indicating that 60% of the root tissue faces resistance; if the root density is lower than 1.5 grams and the hardening gradient is lower than 0.4, the inhibition intensity is 0.1.In addition, the respiration inhibition gradient is normalized by multiplying the oxygen permeability coefficient per unit root density by the ratio of the compaction degree. Taking the soil layer of 20 - 30 cm at point M - 45 as an example, the root density is 2.8 g per cubic decimeter, the corresponding compaction gradient is 0.65, the set oxygen permeability coefficient is 1.3 mg per hour per cubic decimeter, and the compaction degree is 1.5 MPa. Then the normalized result is 0.43. Finally, the root growth inhibition gradient and the respiration inhibition gradient are linearly averaged to form the comprehensive constraint gradient value at this point. This operation is calculated in parallel for all grid cells and depth layers to generate a complete three - dimensional gradient structure for input to the next - stage root absorption analysis. For the operation of step S246 "Estimate the root absorption limitation of the predicted data of the seedling root density based on the nutrient conversion inhibition fluctuation value, the root growth / respiration inhibition gradient data, and the soil nutrient imbalance effectiveness loss data to obtain the root absorption limitation estimation data", it is necessary to construct a main index based on the predicted data of the seedling root density, and pair - integrate the nutrient conversion inhibition fluctuation value (unitless) generated from step S243, the root growth and respiration inhibition gradient (value range from 0 to 1) generated from step S245, and the soil nutrient imbalance effectiveness loss data (expressed as the effectiveness loss ratio, unit is percentage) according to the root - dense unit. In the operation, the root - dense unit is used as the minimum estimation unit. First, the three - item data of this unit are normalized, and the standard interval is set to [0,1]. Among them, the maximum value of the conversion inhibition value is set to 10, and the maximum value of the imbalance loss ratio is set to 90%. Then, the weighted linear combination method is used to assign weights for the limited - value estimation. The weights are 0.3 for the conversion inhibition item, 0.5 for the growth / respiration constraint item, and 0.2 for the imbalance loss item. For example, in the 20 - 30 cm root layer of area M - 45, the predicted root density value is 2.4 g per cubic decimeter, the corresponding conversion inhibition fluctuation value is 7.8, and after normalization, it is 0.78; the corresponding growth and respiration constraint gradient is 0.63; the soil effectiveness loss ratio is 58%, and after normalization, it is 0.644. The comprehensive weighted value is 0.7154, which represents the relative absorption limitation degree of the roots in this unit under the current conditions. The higher the value, the more serious the limitation. Finally, the estimated values of all root - dense units are filled back into the three - dimensional root distribution grid to form a complete root absorption limitation estimation data structure for triggering the judgment condition of the subsequent fertilization and watering fine - tuning control instructions. This operation does not involve the simulation process throughout, and only relies on the original data and the logical weight relationship to achieve the analysis of the limitation degree.

[0125] Step S3 includes the following steps:

[0126] Step S31: Match the slow - release fertilizer application rate control data according to the root absorption limitation estimation data for the abnormal fluctuation data of the soil salt content to obtain the slow - release fertilizer application rate control data;

[0127] Step S32: Conduct usage control logic learning on the slow-release fertilizer usage control data to obtain slow-release fertilizer usage control learning data;

[0128] Step S33: Based on the slow-release fertilizer usage control learning data and the root absorption limitation estimation data, perform segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio for fertilization and watering.

[0129] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0130] Step S31: According to the root absorption limitation estimation data, perform slow-release fertilizer usage control matching on the abnormal fluctuation data of soil salt content to obtain slow-release fertilizer usage control data;

[0131] In the embodiment of the present invention, for the operation of step S31 "According to the root absorption limitation estimation data, perform slow-release fertilizer usage control matching on the abnormal fluctuation data of soil salt content to obtain slow-release fertilizer usage control data", first, a corresponding relationship needs to be established between the root absorption limitation estimation data and the abnormal fluctuation data of soil salt content according to the spatial position. The root absorption limitation estimation data comes from step S246, with the unit being a dimensionless absorption limitation coefficient (between 0 and 1), and the abnormal fluctuation data of soil salt content comes from step S241, which is the standard deviation of salt within 7 days in the soil per unit volume, with the unit being grams per kilogram. During the operation, a basic matching unit is a soil unit of 0.5 m × 0.5 m × 0.3 m. First, the initial proportion limit of the slow-release fertilizer usage is determined according to the absorption limitation estimate. When the maximum limit coefficient is 1.0, the slow-release fertilizer is completely suspended, and when the minimum limit coefficient is 0, the basic fertilization amount is maintained. Let the basic fertilization amount be 6 grams of slow-release nitrogen-potassium mixed particles per unit per day. Subsequently, the basic value is adjusted according to the soil salt fluctuation range. If the fluctuation value exceeds 3.0 grams per kilogram, it is reduced by 30% based on the basic fertilization amount. If it is lower than 1.0 grams per kilogram, it is not adjusted. If it is in the intermediate range, it is linearly decreased proportionally. The specific adjustment ratio is obtained by subtracting 1.0 from the current fluctuation value, dividing by 2.0, and then multiplying by 0.3 to get the reduction ratio. For example, in the nursery area of unit A-12, the absorption limitation estimate is 0.67, and the salt fluctuation is 2.3 grams per kilogram. The corresponding basic fertilization amount is 6 grams. According to the limit coefficient reduction ratio of 0.67, the preliminary fertilization amount is 6 multiplied by (1 - 0.67) to get 1.98 grams. Then, combined with the salt fluctuation adjustment amount, since the fluctuation value is between 1.0 and 3.0, it is calculated as (2.3 - 1.0) divided by 2.0 and then multiplied by 0.3, and the result is 0.195, that is, a further reduction of 19.5%. The final slow-release fertilizer application amount is 1.98 multiplied by (1 - 0.195) to get approximately 1.59 grams. This operation is performed synchronously on all monitoring units to form a complete slow-release fertilizer usage control data grid and output it for subsequent learning and fusion.

[0132] Step S32: Perform dosage control logic learning on the slow-release fertilizer dosage control data to obtain slow-release fertilizer dosage control learning data;

[0133] In the embodiment of the present invention, for the operation of step S32 "Perform dosage control logic learning on the slow-release fertilizer dosage control data to obtain slow-release fertilizer dosage control learning data", it is necessary to perform sliding analysis and dynamic adjustment factor extraction on the slow-release fertilizer dosage control data in the time series dimension to form regular control logic data. The operation steps include data normalization processing, trend factor extraction, upper and lower fluctuation threshold determination, local abnormal increase identification, and gradient change feature statistics. In the specific implementation process, first perform daily normalization processing on the slow-release fertilizer dosage control data for 14 consecutive days, with the standardization range being 6 grams for the daily fertilization upper limit and 0 grams for the lower limit. Subsequently, set the period to 5 days in a sliding window manner, and extract the maximum value, minimum value, mean value, and range within each 5-day window to judge the local fluctuation trend. Set the fluctuation determination threshold that if the range exceeds 0.4 of the normalized value, it is marked as an unstable section. If two unstable sections appear continuously, then insert the logic mark "fluctuation overload" at this position. Taking the area numbered C-9 as an example, the normalized values from the 1st to the 5th day are [0.30, 0.36, 0.42, 0.50, 0.46], and the range is 0.20, which is less than the threshold 0.4 and is not marked as unstable; from the 6th to the 10th day, it is [0.41, 0.67, 0.72, 0.49, 0.75], and the range is 0.34, which still does not exceed; but from the 11th to the 15th day, it is [0.35, 0.89, 0.94, 0.81, 0.87], and the range is 0.59, which exceeds the threshold and is marked as an unstable section. Through the above method, the time series fertilization data of all spatial units is transformed into a data structure containing trend, stability, and abnormality labels, and this structure is the slow-release fertilizer dosage control learning data, which is used to support the next-step drip irrigation ratio structure reorganization.

[0134] Step S33: Based on the slow-release fertilizer dosage control learning data and the root absorption limitation estimation data, perform segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio for fertilization and watering.

[0135] In the embodiments of the present invention, for the operation of step S33, "performing segmented drip irrigation ratio fusion based on the slow-release fertilizer application rate control learning data and the root absorption limitation estimation data to obtain the segmented drip irrigation ratio for fertilization and watering", it is necessary to construct a segmented control logic table, and jointly perform hierarchical mapping on the fluctuation trend sections extracted from the slow-release fertilizer application rate control learning data and the absorption limitation intensity in the root absorption limitation estimation data to determine the drip irrigation ratio for each control time slice. In the operation, the daily drip irrigation process is divided into three time periods: 06:00 to 10:00, 12:00 to 15:00, and 17:00 to 20:00, and the initial irrigation ratios for the three periods are respectively 20%, 30%, and 50%. For the units with an absorption limitation intensity in the range of 0.0 to 0.3, no adjustment is made. For the units with an intensity in the range of 0.3 to 0.6, the ratios for the morning and evening sections are each reduced by 10%, and the ratio for the middle section is increased by 20%. For the units with an intensity higher than 0.6, the ratios for the morning and evening are each reduced by 15%, and the ratio for the middle section is increased by 30%. Subsequently, according to the fluctuation trend marks in the learning data, further fine-tuning is performed. If there are more than two consecutive unstable marks, the total daily drip irrigation amount is reduced by 20%, and it is redistributed proportionally among each time period. Taking the monitoring unit numbered B-6 as an example, the root absorption limitation valuation is 0.71, and the control learning data shows "fluctuation overload" marks in two consecutive sliding periods. The initially set segmented ratio adjustments are as follows: for the morning section, 20% minus 15% is 5%; for the middle section, 30% plus 30% is 60%; for the evening section, 50% minus 15% is 35%, and the total is 100%. Then, after uniformly compressing by 20%, the total drip irrigation amount is 80% of the original drip amount. So, the actual morning section is 4%, the middle section is 48%, and the evening section is 28%. Finally, after this operation is performed on all units, a complete spatial-temporal segmented drip irrigation ratio data set is formed as the command parameter data for the terminal drip irrigation control device.

[0136] Step S33 includes the following steps:

[0137] Step S331: Performing respiratory conversion demand space deduction based on the root absorption limitation estimation data to obtain the root respiratory conversion demand space;

[0138] Step S332: Performing nutrient absorption rate interval calculation per unit time on the root absorption limitation estimation data to obtain the root nutrient absorption rate interval per unit time;

[0139] Step S333: Performing fertilizer efficiency release rate analysis on the slow-release fertilizer application rate control learning data to obtain the fertilizer efficiency application release rate;

[0140] Step S334: Performing extreme point splitting processing on the fertilizer efficiency application release rate per unit time according to the root respiratory conversion demand space and the root nutrient absorption rate interval to obtain the root fertilizer efficiency release absorption extreme value data;

[0141] Step S335: Analyze the absorption and release load ratio of the extreme data of root system fertilizer efficiency release and absorption to obtain the absorption and release load ratio of root system fertilizer efficiency;

[0142] Step S336: Based on the absorption and release load ratio of root system fertilizer efficiency, the root system nutrient absorption rate interval, and the root respiration conversion demand space, conduct segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio of fertilization and watering.

[0143] In the embodiment of the present invention, for the operation of step S331 "deducing the respiration conversion demand space based on the root absorption limitation estimation data to obtain the root respiration conversion demand space", first, the root absorption limitation estimation data is rasterized in the form of a two-dimensional spatial grid. The grid cell is 0.5 m × 0.5 m, and the absorption limitation valuation of each grid cell is between 0.0 and 1.0. The higher the value, the more serious the root absorption limitation. In this operation, the basic root respiration oxygen demand is set to 45 grams of oxygen per cubic meter of root zone per day. The basic amount is modulated by the absorption limitation valuation, and a linear correction method is adopted. Let the absorption limitation value be α, then the oxygen demand per unit volume of the root system is the basic oxygen demand multiplied by 1 plus the square of α, and the oxygen demand intensity at different positions is obtained. Subsequently, based on the spatial distribution map, the oxygen demand intensities of all grid cells are interpolated and reconstructed with high-density contour lines. The discrete bidirectional B-spline interpolation method is used to perform Laplace interpolation reconstruction on the continuous space of the oxygen demand distribution, and the root distribution depth layer is distinguished by combining the seedling type information. For deciduous trees, three depth layers are set, and for evergreen shrubs, two depth layers are set, with each layer thickness of 0.2 m. The root respiration conversion demand space is constructed in a three-dimensional coordinate system, and the output structure is a three-dimensional tensor array. The axes correspond to the X and Y plane coordinates and the depth direction respectively, and the tensor value represents the oxygen demand density in the unit volume space. Taking a nursery unit numbered Z-7 as an example, the average absorption limitation valuation in this area is 0.53, and the basic oxygen demand is 45 grams. After modulation, it is 45 multiplied by 1 plus the square of 0.53, and the oxygen demand is approximately 57.65 grams. Through this method, a root respiration conversion demand space based on the soil spatial structure and with the oxygen demand density as the index is completely established for subsequent drip irrigation control space mapping. For the operation of step S332 "calculating the nutrient absorption rate interval within a unit time for the root absorption limitation estimation data to obtain the root nutrient absorption rate interval within a unit time", first, a standardized root activity function is constructed based on the root absorption limitation estimation data. The function takes the absorption limitation valuation α as the input and corresponds to the reduction coefficients of the lower and upper limits of the root absorption rate per unit time within the interval [0, 1]. Taking the total daily fertilization amount as the reference benchmark, the standard maximum absorption rates are defined as 0.45 grams of nitrogen element, 0.38 grams of potassium element, and 0.22 grams of phosphorus element per square meter of root zone per day, respectively, as the maximum absorption rates in the unrestricted state. In the specific operation, according to the absorption limitation valuation, the nitrogen, phosphorus, and potassium absorption capacities of each spatial unit are scaled to the maximum absorption rate multiplied by (1 - α), and at the same time, the lower limit value of the absorption rate interval is generated as the maximum absorption rate multiplied by (1 - α - 0.1), and the lower limit is not less than zero. When α is greater than 0.9, the absorption rate interval is directly assigned as [0, 0]. All data is stored in a three-dimensional array in a structured manner, and the array dimensions correspond to the spatial position, element type, and upper and lower limit values.Taking the nursery unit in area D-5 with the number as an example, the absorption-limited valuation of this area is 0.37. Then the upper limit of nitrogen element absorption is 0.45 multiplied by (1 minus 0.37), which is approximately 0.2835 grams, and the lower limit is 0.45 multiplied by (1 minus 0.47), which is approximately 0.2385 grams; for potassium element, it is 0.38 multiplied by (1 minus 0.37), which is approximately 0.2394 grams, and the lower limit is 0.38 multiplied by (1 minus 0.47), which is approximately 0.2014 grams; for phosphorus element, it is 0.22 multiplied by (1 minus 0.37), which is approximately 0.1386 grams, and the lower limit is 0.22 multiplied by (1 minus 0.47), which is approximately 0.1166 grams. Through the above calculations, a complete three-dimensional absorption rate interval of space-time-nutrient is formed for setting the boundary constraints of fertilization control ratio and drip irrigation frequency. For the operation of step S333 "Analyze the fertilizer efficiency release rate of the learning data on the controlled amount of slow-release fertilizer to obtain the release rate of fertilizer efficiency dosage", first, based on the daily application amount in the learning data on the controlled amount of slow-release fertilizer, the time-sharing release rate is deduced by combining the fertilization period and soil environment data. The release of slow-release fertilizer is mainly controlled by soil temperature, soil humidity and microbial activity. During the implementation, the daily environmental parameters are synchronously sampled and recorded. The sampling time is set once every two hours, and the sampling parameters include the temperature (unit: degree Celsius) and water content (unit: volume percentage) at 5 cm below the ground surface, and the soil microbial activity index is introduced as an auxiliary factor. The basic release model is set as a piecewise linear release model. When the soil temperature is 25 degrees Celsius, the humidity is 22%, and the microbial index is medium, the 24-hour release rate of each gram of slow-release fertilizer is 18%. Under different environmental conditions, the release coefficient is corrected according to the following adjustment rules: the release rate increases by 1.2 percentage points for every 1 degree Celsius increase in temperature, 0.6 percentage points for every 1 percentage point increase in humidity, and 1.5 percentage points for every increase in one level of microbial activity (divided into weak, medium, strong). Taking the area numbered F-4 as an example, when the daily average temperature is 27 degrees Celsius, the average humidity is 24%, and the microbial activity is "strong", the increase brought by temperature is 1.2 multiplied by 2, which is 2.4 percentage points, the increase brought by humidity is 0.6 multiplied by 2, which is 1.2 percentage points, and the increase brought by the microbial factor is 1.5 percentage points. The total release rate is 18 plus 2.4 plus 1.2 plus 1.5, which is equal to 23.1%. If the daily fertilization amount of this unit is 2.5 grams, then the daily release amount is 2.5 multiplied by 23.1%, which is approximately 0.5775 grams. In this way, the cumulative release of the fertilization amount in all time periods is calculated, a time series of the release amount every two hours within 24 hours is constructed, the release rate of fertilizer efficiency dosage is summarized and output, and a structured time series matrix is formed for the subsequent dynamic coordination of drip irrigation flow and fertilization amount.

[0144] For the operation of step S334, "Performing the extreme point slicing process of the fertilizer efficiency dosage release rate per unit time according to the root respiration conversion demand space and the root nutrient absorption rate interval to obtain the root fertilizer efficiency release and absorption extreme value data", first, perform the synchronous registration process of the root respiration conversion demand space tensor and the root nutrient absorption rate interval tensor per unit time in the spatial dimension, use a unified three-dimensional space grid system as the index structure, set the grid size to 0.25 m × 0.25 m × 0.2 m, and uniformly interpolate all tensors to this spatial resolution. Then, perform the time synchronous slicing process on the fertilizer efficiency dosage release rate time series matrix, perform sliding slicing with one hour as the time step, compare the release rate value within each time step with the upper limit value of the absorption rate in this area of the root system point by point, extract the intersection time points where the difference is lower than the set threshold, and set the threshold as the fertilizer efficiency release rate being less than or equal to the nutrient absorption upper limit value plus 5%. At the same time, according to the root respiration conversion oxygen demand density value, perform weighted screening on the above intersection time points, and only retain the release-absorption intersection points in the area where the oxygen demand density is greater than 50 grams per cubic meter per day, and consider that the root activity in this area meets the high absorption requirements. Organize all the screening results into a structured four-dimensional array, with dimensions including the X coordinate, Y coordinate, depth index, and time index. The array value records the release rate, absorption rate, and oxygen demand density corresponding to the extreme absorption time point, which is defined as the root fertilizer efficiency release and absorption extreme value data. Taking the unit area numbered F-12 in area F as an example, at the third layer depth position, the nitrogen release rate is 0.12 grams per hour, the absorption upper limit is 0.13 grams per hour, and the oxygen demand density is 62 grams per cubic meter, meeting all the conditions. Then this time point is recorded as an extreme point, further record that the extreme time step is the 6th hour, and the corresponding coordinates are X = 6.25, Y = 8.00, and the depth index is 2.

[0145] For the operation of step S335 "Analyze the absorption and release load ratio of the extreme data of root system fertilizer efficiency release and absorption to obtain the absorption and release load ratio of root system fertilizer efficiency", first, the extreme data extracted in step S334 is processed by element, and separate analysis is carried out according to three nutrient categories of nitrogen, phosphorus, and potassium. For each extreme data point, the absorption rate and release rate values are extracted, and the absorption and release load ratio is calculated by the direct ratio method, that is, the absorption rate is used as the numerator and the release rate is used as the denominator to obtain the absorption and release load ratio of the spatial unit per unit time. To avoid the disturbance of extreme data, all data points with a release rate less than 0.01 grams per hour are excluded. All ratio results are reconstructed in the spatial dimension to generate a three-dimensional spatial field, where the dimensions correspond to the X, Y, and depth indices, and the tensor value is the average absorption and release load ratio at the corresponding position. At the same time, the extreme value interval is recorded, and the maximum and minimum thresholds are set in the extreme value distribution area. The points with a ratio greater than 1.5 or less than 0.3 are set as high-load or low-load areas and marked in a separate logical mask matrix. Taking the area with nursery number E-9 as an example, at the corresponding position X = 10.5, Y = 5.25, and depth index 1 at the 4th hour, the absorption rate is 0.18 grams per hour, the release rate is 0.15 grams per hour, and the calculated load ratio is 0.18 divided by 0.15, which is 1.2, falling into the normal load range. Taking the area with number H-3 as another example, the absorption rate is 0.08 grams, the release rate is 0.2 grams, and the ratio is 0.4, which is determined as a low-load area, and the value of this point is set to "L" in the mask matrix. Through this method, a structured absorption and release load ratio data field is formed, providing a quantitative index support for the next step of ratio fusion. For the operation of step S336 "Based on the absorption and release load ratio of root system fertilizer efficiency, the root system nutrient absorption rate interval, and the root system respiration conversion demand space, perform segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio of fertilization and watering", first, a unified spatial data framework is established, and the absorption and release load ratio tensor, nutrient absorption rate interval tensor, and respiration conversion demand space tensor are uniformly mapped to the same spatial grid system, with each grid dimension being 0.25 m × 0.25 m × 0.2 m, and the spatial positions of the three groups of data are aligned. Subsequently, three types of regulation factors are defined: load factor, absorption factor, and aerobic factor, which are determined by the load ratio value, the upper limit value of the absorption rate, and the aerobic density value respectively. For each spatial unit, the regulation weight coefficients are set to 0.4, 0.35, and 0.25 respectively, and a weighted linear combination method is used to fuse and generate the drip irrigation regulation index per unit space. The value range of the drip irrigation regulation index is between 0 and 1, and the higher the value, the higher the required fertilization drip irrigation ratio. Based on this index value, it is divided into five grade intervals, and the drip irrigation ratios are set to 0.6, 0.75, 0.9, 1.05, and 1.2 respectively.Taking the cell in the area numbered G-6 as an example, the load ratio is 1.15, the upper limit of the absorption rate is 0.14 grams per hour, and the aerobic density is 55 grams per cubic meter. Then the regulation index is 1.15 multiplied by 0.4 plus 0.14 multiplied by 0.35 plus 0.055 multiplied by 0.25, and the result is approximately 0.689, falling into the range of 0.6 to 0.75. The corresponding drip irrigation ratio is set to 0.75 times the reference drip irrigation flow rate. The reference flow rate is 2 liters per hour, so the final drip irrigation ratio of this unit is 1.5 liters per hour. The drip irrigation ratios of all units form a three-dimensional space structure matrix, which is matched and mapped in combination with the numbers of the ground drip irrigation execution devices to generate a spatial drip irrigation ratio control map, and finally used to generate a segmented fertilization and watering control instruction sequence.

[0146] Step S4 includes the following steps:

[0147] Step S41: Normalize the segmented drip irrigation ratio of fertilization and watering to obtain the normalized segmented drip irrigation ratio of fertilization and watering;

[0148] Step S42: Based on the policy gradient algorithm, construct a fertilization and watering control model for the normalized segmented drip irrigation ratio of fertilization and watering to obtain a fertilization and watering control model for landscape nursery stock;

[0149] Step S43: Send the fertilization and watering control model for landscape nursery stock to the terminal to execute the automatic fertilization and watering control of landscape nursery stock.

[0150] In the embodiment of the present invention, for the specific operation of step S41 "normalize the segmented drip irrigation ratio of fertilization and watering to obtain the normalized ratio of segmented drip irrigation of fertilization and watering", first collect the three-dimensional spatial structured drip irrigation ratio matrix data generated by step S336. This matrix covers all drip irrigation cell points within the coverage area, and its drip irrigation ratio value is the value weighted and fused according to the load ratio, oxygen demand density, and absorption rate, with the unit of liters per hour. To ensure unified scheduling and management of the drip irrigation ratios in each area under the same control scale, set a normalization operation benchmark. Taking the maximum and minimum values among all drip irrigation ratio values in the current monitoring period as the boundaries, use linear normalization to map all drip irrigation ratio values to the [0,1] interval. The normalization method is: subtract the global minimum value from each unit drip irrigation ratio, and then divide by the difference between the maximum value and the minimum value. Taking the cell in the area numbered F-8 as an example, if its original drip irrigation ratio is 1.05 liters per hour, the minimum drip irrigation ratio in the current area is 0.6 liters per hour, and the maximum drip irrigation ratio is 1.2 liters per hour, then the normalized value is (1.05 minus 0.6) divided by (1.2 minus 0.6), and the result is 0.75. All normalized values retain their three-dimensional grid index structure according to their spatial positions unchanged, and finally form a normalized drip irrigation ratio spatial tensor. The tensor dimension is exactly the same as the original drip irrigation ratio matrix, serving as the input data basis for subsequent strategy optimization and control model construction. All normalized numerical values are retained to three decimal places to ensure accuracy consistency. All calculation processes are carried out in a fixed parameter control environment, and the normalization results are written into an independent control parameter structure body and bound to the independent identification code of each control area to ensure the integrity of the corresponding relationship in the subsequent model construction process.

[0151] For the specific operation of step S42, "Construct a fertilization and watering control model for the segmented drip irrigation normalization ratio of garden seedlings based on the policy gradient algorithm to obtain a fertilization and watering control model for garden seedlings", the construction of the control strategy is realized by using the policy gradient algorithm based on the advantage function. First, the drip irrigation normalization ratio tensor generated in step S41 is used as the state input, combined with the historical operation execution record data and the soil moisture content change data as the feedback data set, to establish a state-behavior-feedback triple data set. The state space is defined as the normalized drip irrigation ratio tensor at the current moment, the behavior space is defined as a two-dimensional vector composed of the drip irrigation opening time and the fertilization concentration of each cell, and the feedback is the product function value of the change value of the soil moisture content and the change value of the root activity in this area within two hours. For this triple, the behavior strategy function is designed to construct a probability strategy function in the form of a Gaussian distribution, and the current strategy is enhanced using the behavior advantage function. The advantage function is composed of the return value minus the expected value of the state value function, where the state value function is obtained by fitting historical feedback data. The Monte Carlo sampling method is used to generate the state-behavior paths updated in each round, and the sampling quantity is set to 1000 trajectories per round. In each iteration, according to the policy gradient formula, the gradient direction of the current strategy under the normalization ratio is calculated, and the Adam optimizer with a learning rate of 0.001 is used for gradient ascent operation. The parameters of the strategy function are updated once after each round of iteration, and the number of iteration rounds is set to 500 rounds. All training processes are executed under the CPU-GPU collaborative architecture, and the state tensors are sent to the memory in batches, with the single-round calculation time controlled within 20 seconds. Taking the area numbered G-10 as an example, in the 200th round of training, its drip irrigation normalization ratio is 0.78, and the optimal behavior output by the policy network is to turn on the drip irrigation for 7 minutes with a fertilization concentration of 0.06 grams per liter. Finally, all state-behavior mapping relationships are sorted into a set of continuous function expressions, and a control mapping function group is established under the corresponding spatial index to form a parameter set of the fertilization and watering control model for garden seedlings. For the specific operation of step S43, "Send the fertilization and watering control model for garden seedlings to the terminal to execute the automatic fertilization and watering control of landscape garden seedlings", first, the control strategy data structures corresponding to each cell in the fertilization and watering control model constructed in step S42 are packed, and the format is the binary compression structure format. The file structure includes six items: area number, three-dimensional coordinate index, normalization ratio, corresponding fertilization concentration, drip irrigation opening duration, and strategy evaluation score. All structures are written into the batch number T-SFQ-0327. To ensure data compatibility with the terminal control system, the encoding of all structure fields follows the preset control protocol standard, and the numerical fields are represented by 32-bit floating-point numbers, and the coordinate index field is encoded as an 8-bit integer. After the data packing is completed, it is sent to the control receiving module corresponding to each terminal number in groups through the standard local area network communication module using the UDP communication protocol. The communication port number is set to DIO-13, and the communication message period is set to be updated every 120 seconds.After receiving the policy model data packet, the terminal device calls the drip irrigation and fertilization control driver in the local control execution module, restores the received normalization ratio to the specific drip irrigation flow rate, and controls the opening rate of the pump pressure and the fertilizer mixing solenoid valve according to the fertilization concentration and the opening duration. Taking the seedling area number D-4 as an example, the fertilization concentration in the received policy control structure is 0.07 grams per liter, and the opening duration is 8 minutes. The terminal controller calculates the total drip irrigation flow rate to be 1.6 liters and the total fertilization amount to be 0.112 grams based on this data. The system execution cycle completes all operation processes within 1 minute after data reception, and records the execution status code and uploads it to the control center for closed-loop control status feedback. Through the above method, the deployment of the control model at the terminal and the precise implementation of the automated fertilization and watering operation are completed.

[0152] The present invention also provides a landscape nursery stock automated fertilization and watering control system for implementing the landscape nursery stock automated fertilization and watering control method as described above. The landscape nursery stock automated fertilization and watering control system includes:

[0153] A soil state characteristic monitoring module for collecting nursery stock morphological data of landscape nursery stock through electronic monitoring to obtain nursery stock morphological data; deploying soil monitoring sensors under each landscape nursery stock and monitoring soil state characteristics through the soil detection sensors to obtain soil state characteristic monitoring feature data;

[0154] A root absorption limitation estimation module for predicting the density of nursery stock roots based on the nursery stock morphological data to generate nursery stock root density prediction data; analyzing the abnormal fluctuations of soil salt content in the soil state characteristic monitoring feature data to obtain soil salt content abnormal fluctuation data; estimating the root absorption limitation of the nursery stock root density prediction data based on the soil salt content abnormal fluctuation data to obtain root absorption limitation estimation data;

[0155] A segmented drip irrigation ratio fusion module for controlling and matching the slow-release fertilizer dosage according to the root absorption limitation estimation data for the soil salt content abnormal fluctuation data to obtain slow-release fertilizer dosage control data; performing segmented drip irrigation ratio fusion based on the slow-release fertilizer dosage control data to obtain a segmented drip irrigation ratio for fertilization and watering;

[0156] A fertilization and watering control model construction module for constructing a fertilization and watering control model for landscape nursery stock based on the policy gradient algorithm for the segmented drip irrigation ratio for fertilization and watering to obtain a fertilization and watering control model for landscape nursery stock; sending the fertilization and watering control model for landscape nursery stock to the terminal to execute the landscape nursery stock automated fertilization and watering control.

[0157] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. An automatic fertilization and watering control method for landscape nursery stock, characterized in that, It includes the following steps: Step S1: Collect seedling morphological data of landscape nursery stock through electronic monitoring to obtain seedling morphological data; Deploy soil monitoring sensors under each landscape nursery stock, and monitor the soil state characteristics through the soil detection sensors to obtain soil state characteristic monitoring data; Step S2: Predict the root density of the nursery stock based on the seedling morphological data to generate root density prediction data of the nursery stock; Analyze the abnormal fluctuations of soil salt content in the soil state characteristic monitoring data to obtain abnormal soil salt content fluctuation data; Estimate the root absorption limitation of the root density prediction data of the nursery stock according to the abnormal soil salt content fluctuation data to obtain root absorption limitation estimation data; Step S3: Control and match the slow-release fertilizer dosage according to the root absorption limitation estimation data for the abnormal soil salt content fluctuation data to obtain slow-release fertilizer dosage control data; Fuse the segmented drip irrigation ratio based on the slow-release fertilizer dosage control data to obtain the segmented drip irrigation ratio for fertilization and watering; Step S4: Construct a fertilization and watering control model for the segmented drip irrigation ratio of fertilization and watering based on the policy gradient algorithm to obtain a fertilization and watering control model for landscape nursery stock; Send the fertilization and watering control model for landscape nursery stock to the terminal to perform automatic fertilization and watering control of landscape nursery stock.

2. The automated fertilization and watering control method for landscape nursery stock according to claim 1, characterized in that Step S1 includes the following steps: Step S11: Collect seedling morphological data of landscape nursery stock through electronic monitoring to obtain seedling morphological data; Step S12: Deploy soil monitoring sensors under each landscape nursery stock, and monitor the soil state characteristics through the soil detection sensors to obtain soil state characteristic monitoring data; Step S13: Clean the soil state characteristic monitoring data to obtain cleaned soil state characteristic monitoring data; Step S14: Analyze the state characteristic features of the cleaned soil state characteristic monitoring data to obtain soil state characteristic monitoring data.

3. The automated fertilization and watering control method for landscape nursery stock according to claim 1, wherein, Step S2 includes the following steps: Step S21: Predict the root density of the nursery stock based on the seedling morphological data to generate root density prediction data of the nursery stock; Step S22: Analyze the abnormal fluctuations of soil salt content in the soil state characteristic monitoring data to obtain abnormal soil salt content fluctuation data; Step S23: Analyze the effectiveness loss of soil nutrient imbalance in the soil state characteristic monitoring data based on the abnormal soil salt content fluctuation data to obtain soil nutrient imbalance effectiveness loss data; Step S24: Estimate the root absorption limitation of the root density prediction data of the nursery stock according to the abnormal soil salt content fluctuation data and the soil nutrient imbalance effectiveness loss data to obtain root absorption limitation estimation data.

4. The automated fertilization and watering control method for landscape nursery stock according to claim 3, characterized in that, Step S23 includes the following steps: Step S231: Decompose and process the abnormal soil salt content fluctuation data to generate time series change data of salt ion concentration; Step S232: Extract the fluctuation of organic matter content in the soil state characteristic monitoring data to obtain soil organic matter content fluctuation data; Step S233: Perform non-linear strength regression analysis of cell membrane damage based on the time series change data of salt ion concentration to obtain cell membrane damage strength regression data; Step S234: Perform an equivalent simulation prediction of organic matter nutrient leakage on the soil organic matter content fluctuation data based on the cell membrane damage intensity regression data to obtain equivalent prediction data of nutrient leakage; Step S235: Analyze the increment of ion competition intensity on the soil organic matter content fluctuation data according to the time-series change data of salt ion concentration to generate ion competition intensity increment data; Step S236: Conduct an analysis of the effective loss of soil nutrient imbalance based on the equivalent prediction data of nutrient leakage and the ion competition intensity increment data to obtain soil nutrient imbalance effective loss data.

5. The automated fertilization and watering control method for landscape nursery stock according to claim 4, characterized in that, Step S235 includes the following steps: Perform a mutation increment index calculation on the time-series change data of salt ion concentration to obtain a salt ion concentration mutation increment index; Based on the salt ion concentration mutation increment index, perform a continuous concentration mutation approximation combination on the time-series change data of salt ion concentration to generate concentration continuous mutation approximation combination data; Based on the concentration continuous mutation approximation combination data, perform a preemptive continuous intensity simulation of salt ion absorption sites on the soil organic matter content fluctuation data to obtain preemptive continuous intensity data of salt ion absorption sites; Perform an ion flux time-series interval difference calculation on the preemptive continuous intensity data of absorption sites to obtain the salt ion flux difference at the absorption sites; Analyze the increment of ion competition intensity according to the salt ion flux difference at the absorption sites to generate ion competition intensity increment data.

6. The automated fertilization and watering control method for landscape nursery stock according to claim 3, characterized in that, Step S24 includes the following steps: Step S241: Perform a fluctuation amplitude deviation rate calculation on the abnormal fluctuation data of soil salt content to obtain salt content fluctuation amplitude difference data; Step S242: Convert and quantify the increase ratio of soil solution osmotic pressure according to the salt content fluctuation amplitude difference data to obtain the increase ratio of soil solution osmotic pressure; Step S243: Perform a numerical fluctuation integration of microbial nutrient transformation inhibition based on the increase ratio of soil solution osmotic pressure to obtain a nutrient transformation inhibition fluctuation value; Step S244: Estimate the cumulative gradient of soil compaction according to the salt content fluctuation amplitude difference data and the increase ratio of soil solution osmotic pressure to obtain soil compaction cumulative gradient data; Step S245: Evaluate the root growth / respiration restriction gradient on the predicted data of the density of seedling roots based on the soil compaction cumulative gradient data to obtain root growth / respiration restriction gradient data; Step S246: Estimate the root absorption limitation on the predicted data of the density of seedling roots according to the nutrient transformation inhibition fluctuation value, the root growth / respiration restriction gradient data, and the soil nutrient imbalance effective loss data to obtain root absorption limitation estimation data.

7. The automated fertilization and watering control method for landscape nursery stock according to claim 1, wherein Step S3 includes the following steps: Step S31: Match the controlled release fertilizer dosage for the abnormal fluctuation data of soil salt content according to the root absorption limitation estimation data to obtain controlled release fertilizer dosage control data; Step S32: Learn the dosage control logic for the controlled release fertilizer dosage control data to obtain controlled release fertilizer dosage control learning data; Step S33: Perform a segmented drip irrigation ratio fusion based on the controlled release fertilizer dosage control learning data and the root absorption limitation estimation data to obtain a segmented drip irrigation ratio for fertilization and watering.

8. The automated fertilization and watering control method for landscape nursery stock according to claim 7, characterized in that, Step S33 includes the following steps: Step S331: Based on the estimated data of root absorption limitation, conduct a deduction of the respiration conversion demand space to obtain the root respiration conversion demand space; Step S332: Conduct a calculation of the nutrient absorption rate interval within a unit time for the estimated data of root absorption limitation to obtain the root nutrient absorption rate interval within a unit time; Step S333: Analyze the fertilizer efficiency release rate for the slow-release fertilizer dosage control learning data to obtain the fertilizer efficiency dosage release rate; Step S334: Perform a segmentation process on the fertilizer efficiency dosage release rate based on the root respiration conversion demand space and the root nutrient absorption rate interval to obtain the extreme value data of root fertilizer efficiency release and absorption; Step S335: Analyze the absorption and release load ratio for the extreme value data of root fertilizer efficiency release and absorption to obtain the root fertilizer efficiency absorption and release load ratio; Step S336: Based on the root fertilizer efficiency absorption and release load ratio, the root nutrient absorption rate interval, and the root respiration conversion demand space, conduct a segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio for fertilization and watering.

9. The automated fertilization and watering control method for landscape nursery stock according to claim 1, characterized in that Step S4 includes the following steps: Step S41: Normalize the segmented drip irrigation ratio for fertilization and watering to obtain the normalized segmented drip irrigation ratio for fertilization and watering; Step S42: Based on the policy gradient algorithm, construct a fertilization and watering control model for the normalized segmented drip irrigation ratio for fertilization and watering to obtain the fertilization and watering control model for landscape nursery stock; Step S43: Send the fertilization and watering control model for landscape nursery stock to the terminal to execute the automatic fertilization and watering control for landscape nursery stock.

10. An automatic fertilization and watering control system for landscape nursery stock, characterized in that, For implementing the automatic fertilization and watering control method for landscape nursery stock as described in Claim 1, the automatic fertilization and watering control system for landscape nursery stock includes: A soil state characteristic monitoring module, which is used to collect nursery stock morphological data for landscape nursery stock through electronic monitoring to obtain nursery stock morphological data; deploy soil monitoring sensors under each landscape nursery stock, and conduct soil state characteristic monitoring through the soil detection sensors to obtain soil state characteristic monitoring feature data; A root absorption limitation estimation module, which is used to predict the density of nursery stock roots based on the nursery stock morphological data to generate nursery stock root density prediction data; conduct an analysis of abnormal fluctuations in soil salt content for the soil state characteristic monitoring feature data to obtain abnormal fluctuation data of soil salt content; based on the abnormal fluctuation data of soil salt content, conduct a root absorption limitation estimation on the nursery stock root density prediction data to obtain the estimated data of root absorption limitation; A segmented drip irrigation ratio fusion module, which is used to match the slow-release fertilizer dosage control based on the estimated data of root absorption limitation and the abnormal fluctuation data of soil salt content to obtain the slow-release fertilizer dosage control data; based on the slow-release fertilizer dosage control data, conduct a segmented drip irrigation ratio fusion to obtain the segmented drip irrigation ratio for fertilization and watering; A fertilization and watering control model construction module, which is used to construct a fertilization and watering control model for the segmented drip irrigation ratio for fertilization and watering based on the policy gradient algorithm to obtain the fertilization and watering control model for landscape nursery stock; send the fertilization and watering control model for landscape nursery stock to the terminal to execute the automatic fertilization and watering control for landscape nursery stock.

Images