A smart garden management method and system based on the Internet of Things

Abstract

The application discloses a kind of based on Internet of Things's smart garden management method and system, belong to smart garden technical field.It includes: based on the multiple Internet of Things sensor nodes deployed in garden area, real-time collection plant physiological ecology signal and microclimate environmental data;The plant physiological ecology signal is intercepted by photosynthetically active radiation efficiency analysis, combined with the evapotranspiration parameter in microclimate environmental data, calculate the seasonal physiological response signal of plant community;According to the seasonal physiological response signal, construct the time sequence phenology continuum gradient representing plant phenology evolution;According to the time sequence phenology continuum gradient, dynamically switch horticulture management mode, generate and current phenological stage matched irrigation and pruning operation instruction.The scheme combines the internal physiological rhythm of plant with continuous phenology gradient, realizes the adaptive matching of operation parameter and phenology gradient.

Description

Technical Field

[0001] This invention relates to the fields of smart garden and Internet of Things (IoT) technology, specifically to a smart garden management method and system based on IoT. Background Technology

[0002] The Internet of Things (IoT) technology has been used in garden management for environmental parameter collection and automatic irrigation control. Common methods involve using soil moisture sensors, air temperature and humidity sensors, and weather stations to initiate irrigation when soil moisture falls below a set threshold, while pruning is performed according to fixed cycles or manual experience. Some solutions incorporate phenological calendars, dividing the budding and growth stages based on solar terms or accumulated temperature to adjust irrigation volume or pruning frequency. However, these methods focus only on external environmental factors or broad seasonal divisions, failing to deeply analyze the physiological and ecological signals of the plants themselves. The actual water consumption and growth requirements of a plant community depend on the dynamic relationship between the photosynthetic organ's light capture efficiency and atmospheric transpiration. During seasonal transitions, the photosynthetic electron transfer rate of leaves and the canopy transpiration pull change synergistically; this coupled process determines the plant's true water requirements and growth status. Existing irrigation decisions lack joint quantification of photosynthetically active radiation capture efficiency and transpiration parameters, failing to generate continuous signals characterizing the intensity of the community's intrinsic seasonal physiological response. Therefore, it is difficult to overcome phenological misjudgments and supply-demand mismatches caused by relying solely on soil moisture or temperature and humidity thresholds. Meanwhile, the existing management model divides phenology into several discrete stages, with abrupt transitions and significant changes in irrigation depth and pruning intensity during these stage jumps, ignoring the continuous and gradual changes between budding, vigorous growth, and senescence. Significant differences exist within the microenvironment of the garden area, and this discrete-stage management cannot adapt to the continuous transitional needs of plant community phenological evolution. This leads to delayed irrigation, inappropriate pruning, or damage to branches and trunks, reducing the precision of garden maintenance and the efficiency of resource utilization. Summary of the Invention

[0003] This invention provides a smart garden management method and system based on the Internet of Things, aiming to solve the problems of existing technologies that are difficult to integrate photosynthetically active radiation interception efficiency and evapotranspiration to characterize the seasonal physiological response of plant communities, and the inability of phenological management to continuously adapt to the gradual needs of plants due to the use of discrete stages. By constructing a phenological continuum gradient and dynamically generating irrigation and pruning operation instructions accordingly, the invention achieves refined garden management and phenological self-adaptation.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a smart garden management method based on the Internet of Things, comprising the following steps:

[0005] Based on multiple IoT sensor nodes deployed within the garden area, real-time plant physiological and ecological signals and micro-meteorological environment data are collected. The photosynthetically active radiation interception efficiency of the plant physiological and ecological signals is analyzed, and combined with the evapotranspiration parameters in the micro-meteorological environment data, the seasonal physiological response signal of the plant community is calculated. Based on the seasonal physiological response signal, a temporal phenological continuum gradient characterizing plant phenological evolution is constructed. Based on the temporal phenological continuum gradient, the horticultural management mode is dynamically switched, and irrigation and pruning operation instructions highly matched to the current phenological stage are generated.

[0006] Preferably, the real-time acquisition of plant physiological and ecological signals and microclimate environmental data specifically includes: acquiring multimodal heterogeneous data for each IoT sensor node, wherein the multimodal heterogeneous data includes chlorophyll fluorescence-induced dynamics signals, stem micro-change signals, soil matrix water potential signals, canopy temperature signals, and air temperature, humidity, and wind speed signals; imputing missing values ​​and removing anomalous mutation points from the acquired multimodal heterogeneous data to generate a standardized multi-source time-series data matrix. Through multimodal heterogeneous sensing, plant physiological states and environmental stresses can be captured from multiple dimensions, providing a complete data foundation for subsequent analysis.

[0007] Preferably, calculating the seasonal physiological response signal of a plant community specifically includes: extracting the estimated value of leaf photosynthetic electron transport rate from plant physiological and ecological signals, extracting the air saturated vapor pressure difference and net radiation flux from micrometeorological environmental data, and constructing a plant health index model that integrates seasonal evolution priors; calling the plant health index model to perform nonlinear coupling calculations on the estimated value of leaf photosynthetic electron transport rate, the air saturated vapor pressure difference, and the net radiation flux, and outputting the seasonal physiological response signal of the plant community. As a technical solution of the present invention, the plant health index model adopts an extreme gradient boosting regression algorithm based on seasonal succession parameters, using the estimated value of leaf photosynthetic electron transport rate, the air saturated vapor pressure difference, and the net radiation flux as input variables to predict the health index, thereby accurately quantifying the intensity of physiological activities of the plant community in different seasons.

[0008] Preferably, constructing a temporal phenological continuum gradient characterizing plant phenological evolution specifically includes: detrending and periodic decomposition of the seasonal physiological response signals to separate baseline and fluctuation components, and extracting abrupt change feature points of phenological transition nodes; based on the temporal localization of multiple abrupt change feature points, spatially aggregating and temporally aligning the seasonal physiological response signals of all sensor nodes within the garden area to generate a temporal phenological continuum gradient of the garden's plant community. This gradient continuously characterizes the dynamic process of plants from budding to decline, overcoming the ambiguity of management switching caused by discrete phenological period divisions.

[0009] Furthermore, the dynamic switching of the horticultural management mode generates irrigation and pruning operation instructions that match the current phenological stage. Specifically, this includes: mapping the temporal phenological continuum gradient to a four-stage phenological state space containing dormancy, budding, vigorous growth, and senescence stages, and identifying the current phenological stage of the park; based on the phenological stage classification, retrieving a preset horticultural management rule base, and generating a combination of horticultural management actions including irrigation start / stop thresholds, irrigation depth, pruning frequency, and pruning intensity. This ensures that management decisions are dynamically synchronized with the actual physiological rhythms of the plants, significantly improving the timeliness and relevance of horticultural operations.

[0010] Furthermore, when generating horticultural management action combinations based on phenological stages, the soil moisture consumption rate for five consecutive days is extracted from micrometeorological data. Combined with the slope changes of the temporal phenological continuum gradient, the irrigation start / stop threshold is dynamically updated. Based on the updated irrigation start / stop threshold, a fuzzy logic controller is invoked to adjust the opening of the irrigation valve, ensuring that the irrigation depth precisely matches the root water absorption layer thickness at the current phenological stage. This scheme avoids over-irrigation or under-irrigation caused by fixed thresholds at different phenological stages, achieving adaptive and precise control of irrigation volume.

[0011] Preferably, after generating irrigation and pruning operation instructions, these instructions are also sent to the intelligent irrigation valves and automatic pruning robotic arms, driving each intelligent device to perform differentiated horticultural operations in real time. During the driving process, the operation path of the automatic pruning robotic arm is dynamically planned, and the cutting angle and cutting stroke of the blade are adjusted in real time according to the branch growth rate described in the temporal phenological continuum gradient. At the same time, the intelligent irrigation valves are controlled by variable frequency pulses, and the coverage radius and flow rate of multiple intelligent sprinkler heads are adjusted in real time during irrigation based on the stem micro-change signals in plant physiological and ecological signals. Through the above closed-loop precise execution and dynamic adjustment, the precision and intelligence level of horticultural operations are significantly improved.

[0012] This invention also provides an IoT-based smart garden management system, comprising: a multi-mode heterogeneous data acquisition module for real-time acquisition of plant physiological and ecological signals and micro-meteorological environment data through multiple IoT sensor nodes deployed within the garden area; a seasonal physiological response calculation module for analyzing the photosynthetically active radiation interception efficiency of plant physiological and ecological signals and calculating the seasonal physiological response signal of the plant community by combining the evapotranspiration parameters in the micro-meteorological environment data; a phenological continuum construction module for constructing a temporal phenological continuum gradient characterizing plant phenological evolution based on the seasonal physiological response signal; a management mode switching module for dynamically switching horticultural management modes based on the temporal phenological continuum gradient and generating irrigation and pruning operation instructions matching the current phenological stage; a smart device driving module for sending irrigation and pruning operation instructions to smart irrigation valves and automatic pruning robotic arms to drive them to perform differentiated horticultural operations; and a closed-loop feedback calibration module for receiving operation feedback data transmitted back from smart devices and cross-validating it with plant physiological and ecological signals to trigger the correction of subsequent operation instructions. Through the synergistic effect of its various modules, the system achieves a closed-loop process from perception, analysis, decision-making to execution and feedback, enabling garden management strategies to be continuously optimized based on the real-time physiological state of plants.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0014] This method analyzes the photosynthetically active radiation interception efficiency of plant physiological and ecological signals and integrates evapotranspiration parameters from micrometeorological data. By extracting estimated values ​​of leaf photosynthetic electron transport rate and combining them with air saturation vapor pressure difference and net radiation flux, a nonlinear coupling is performed using a plant health index model incorporating seasonal evolution priors to calculate the seasonal physiological response signals of plant communities. This approach integrates the intrinsic photosynthetic activity of plants and atmospheric evaporation demand into a unified quantitative framework. The obtained signals faithfully reflect the true seasonal physiological rhythms and water vapor exchange dynamics of the community, eliminating phenological misjudgments caused by relying solely on soil moisture thresholds or accumulated temperature. This makes subsequent phenological measurements and operational triggers more biologically relevant, accurately capturing plant physiological turning points even in complex garden scenarios with significant microenvironmental fluctuations.

[0015] Based on seasonal physiological response signals, detrending and periodic decomposition were performed to separate baseline and fluctuation components and extract phenological transition abrupt change feature points. By spatially aggregating and temporally aligning the response signals of all sensor nodes in the entire garden area, a temporal phenological continuum gradient at the garden scale was constructed. This gradient expresses the traditional discrete phenological stages as a smooth and continuous evolution spectrum. After mapping to the four-stage phenological state space of dormancy, budding, vigorous growth, and senescence, the horticultural management mode no longer adopts a step-by-step switching, but rather performs stepless dynamic adjustment based on gradient changes. Soil moisture consumption rates over five consecutive days were extracted from micrometeorological environmental data. Combined with the slope changes of the temporal phenological continuum gradient, irrigation start and stop thresholds were dynamically updated. The opening of irrigation valves was adjusted through a fuzzy logic controller to ensure that the irrigation depth accurately matches the root water absorption layer thickness under the current phenological stage. Simultaneously, the cutting angle and cutting stroke of the automatic pruning robotic arm blades were adjusted in real time based on the branch growth rate described by the gradient. Variable frequency pulse control was performed on the coverage radius and flow rate of the intelligent sprinkler head based on micro-change signals in the stem. This continuum gradient mechanism enables irrigation and pruning parameters to move smoothly with the gradual phenological trend, eliminating physiological stress and water and fertilizer waste caused by sudden changes in management parameters, and achieving precise synchronization between garden maintenance and the continuous phenological rhythm of plants. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of a smart garden management method based on the Internet of Things;

[0018] Figure 2 This is a flowchart of multi-mode heterogeneous data acquisition and preprocessing for IoT sensor nodes;

[0019] Figure 3 This is a flowchart of the construction and prediction process for a plant health index model that integrates prior knowledge of seasonal evolution.

[0020] Figure 4 This is a flowchart of the construction of a population phenological continuum based on iterative decomposition of seasonal physiological response signals and alignment of mutation points;

[0021] Figure 5 This is a flowchart of the dynamic control of irrigation based on the gradient of the plant phenological continuum. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] See Figure 1 This invention provides a smart garden management method based on the Internet of Things (IoT), comprising: real-time collection of plant physiological and ecological signals and micro-meteorological environment data based on multiple IoT sensor nodes deployed within the garden area; analysis of the photosynthetic effective radiation interception efficiency of the plant physiological and ecological signals, and calculation of the seasonal physiological response signal of the plant community by combining the evapotranspiration parameters in the micro-meteorological environment data; construction of a temporal phenological continuum gradient characterizing plant phenological evolution based on the seasonal physiological response signal; and dynamic switching of horticultural management modes based on the temporal phenological continuum gradient, generating irrigation and pruning operation instructions matching the current phenological stage.

[0024] Example 1:

[0025] In specific implementation, please refer to Figure 2 Multimodal heterogeneous data acquisition is performed on each IoT sensor node. Each IoT sensor node integrates a chlorophyll fluorescence sensor, a stem micro-change sensor, a soil matrix water potential sensor, a canopy temperature sensor, and air temperature, humidity, and wind speed sensors. The chlorophyll fluorescence sensor uses a pulse modulation measurement method, applying saturated pulse light after dark adaptation to record the minimum fluorescence. With maximum fluorescence And calculate the maximum photochemical efficiency of optical system II. As a chlorophyll fluorescence-induced kinetic signal, the stem micro-change sensor employs the principle of a linear variable differential transformer, attaching itself to the epidermis of the main stem of the plant with a fixed contact force, and continuously outputs a stem micro-change signal characterizing changes in stem diameter, with a sampling frequency set to once per minute. The soil matrix water potential sensor is deployed in the main root distribution layer, using the principle of balanced dielectric measurement to acquire the soil matrix water potential signal, with a measurement range of [missing information]. to The canopy temperature sensor is a broadband infrared radiometer that measures the brightness and temperature of the plant canopy in a non-contact manner to obtain the canopy temperature signal. Air temperature, humidity, and wind speed signals are acquired by a temperature and humidity sensor and a three-cup anemometer installed in a ventilated radiation shield two meters above the ground. These five signals constitute multimode heterogeneous data. Each IoT sensor node uploads this multimode heterogeneous data to the local edge computing gateway using the LoRa wireless communication protocol, with a unified time-series label.

[0026] In practice, missing values ​​are imputed in the collected multi-mode heterogeneous data. After receiving the multi-mode heterogeneous data, the edge computing gateway first performs integrity verification on the time series of each sensor channel. When a signal gap is detected at a certain sampling time, imputation is performed at the missing position. The imputation process adopts a neighboring historical value imputation method based on time decay weighting. For node numbered... The sensor types collected by the IoT sensor nodes are: signal sequence If timestamp If a signal value is missing at a certain point, the interpolated value is calculated using the following formula. :

[0027]

[0028] in, This indicates the sampling time interval of IoT sensor nodes. This represents the number of neighboring valid historical signal data points involved in the interpolation calculation, with a value of [value missing]. , Indicates corresponding to the first Time decay weights of neighboring historical data points The calculation formula is , The attenuation coefficient has a value of [value missing]. . Represents timestamp The previous The effective signal value for each sampling period. When there are insufficient consecutive valid data points before the missing timestamp. At that time, then Reduced to the actual number of valid data points.

[0029] In practice, abnormal abrupt changes are removed from the signal sequence after missing value imputation. For each sensor type... Node number is Given a signal sequence, construct a sliding window with a sliding window length set to [value]. There are 1 sampling point, and the sliding step size is set to 1. There are 10 sampling points. Within each sliding window, the median of the signal values ​​within the window is calculated. and median absolute deviation For any signal value within the sliding window When satisfied At that time, the judgment These are considered anomalous mutation points. Signal values ​​identified as anomalous mutation points are removed, and the resulting vacancy positions are filled using the aforementioned missing value imputation method.

[0030] In practice, signal sequences that have undergone anomalous mutation removal and refilling, along with normal signal sequences without anomalies, are time-aligned according to a unified time base to form a multi-source time-series data matrix. Each row of the multi-source time-series data matrix corresponds to a sampling time, and each column corresponds to a signal channel of a sensor type from an IoT sensor node. The column vectors are arranged in the following order: chlorophyll fluorescence-induced dynamics signal from the first IoT sensor node, stem micro-change signal from the first IoT sensor node, soil matrix water potential signal from the first IoT sensor node, canopy temperature signal from the first IoT sensor node, air temperature, humidity, and wind speed signals from the first IoT sensor node, and all signal types corresponding to the second IoT sensor node, and so on, until all signal types of all IoT sensor nodes within the park area are represented. For each column of the multi-source time-series data matrix, z-score normalization is performed to ensure that the mean of each column's signal sequence is 1 / 2. Standard deviation is This yields a standardized multi-source time-series data matrix. This standardized multi-source time-series data matrix is ​​stored in the time-series database of the edge computing gateway for use in the seasonal physiological response calculation process.

[0031] Example 2:

[0032] In specific implementation, please refer to Figure 3 The estimated value of leaf photosynthetic electron transport rate was extracted from plant physiological and ecological signals. These signals include chlorophyll fluorescence-induced kinetic signals obtained by a chlorophyll fluorescence sensor at each sampling time, which record the minimum fluorescence after dark adaptation. Maximum fluorescence Steady-state fluorescence under light adaptation Maximum fluorescence under light adaptation The process for extracting the estimated value of the photosynthetic electron transport rate of leaves is as follows: using steady-state fluorescence under light adaptation. Maximum fluorescence under light adaptation The actual photochemical efficiency of optical system II is calculated using the following formula: The estimated photosynthetic electron transport rate of leaves is the product of the actual photochemical efficiency of photosystem II and the measured value of photosynthetically active radiation above the canopy, multiplied by the plant leaf absorption coefficient, which is set at 0.84. This value is based on the average absorption ratio of photosynthetically active radiation by green leaves. The measured value of photosynthetically active radiation above the canopy is synchronously collected by photosynthetically active radiation sensors installed on each IoT sensor node and serves as a component of plant physiological and ecological signals.

[0033] In practice, the air saturated vapor pressure difference and net radiation flux are extracted from the micrometeorological environment data. The micrometeorological environment data includes air temperature, humidity, and wind speed signals; the air temperature value is obtained from the air temperature and humidity signals. and relative humidity value The saturated vapor pressure difference of air is calculated based on the saturated vapor pressure equation. The calculation process is as follows: first calculate the pressure difference at temperature... Saturated water vapor pressure The calculation method uses Magnus's formula. In the formula The unit is Celsius; then, based on relative humidity... Calculate the actual water vapor pressure , Air saturated water vapor pressure difference for The net radiation flux was calculated by combining the canopy temperature and air temperature and humidity signals from the micrometeorological environment data with the surface radiation balance equation. The input quantities used to calculate the net radiation flux included incident shortwave radiation, surface albedo, atmospheric longwave radiation, and longwave radiation emitted from the surface. The incident shortwave radiation was provided by a total radiation sensor installed in an open area of ​​the garden, and the surface albedo was taken as a typical value of 0.23 for vegetation.

[0034] In practical implementation, a plant health index model integrating prior knowledge of seasonal evolution is constructed. This model employs an extreme gradient boosting regression algorithm based on seasonal succession parameters. The core architecture of this algorithm is an additive ensemble model, consisting of multiple sequentially stacked regression trees. The model has a total of... A tree of return, Set to 150. Each regression tree is a weak learner function. subscript This represents the index of the regression tree in the model. The input feature vector is used as the input feature vector. Each regression tree is paired with the input feature vector. The process involves recursively dividing the data into two partitions until a preset maximum depth is reached or the number of samples in a leaf node satisfies the constraints. The weighted sum of all regression tree outputs constitutes the final prediction output of the model.

[0035] In practical implementation, the input feature vector of the plant health index model It consists of five characteristic components: estimated value of leaf photosynthetic electron transport rate Air saturated water vapor pressure difference Net radiative flux First component of seasonal succession parameters and the second component of seasonal succession parameters The first component of seasonal succession parameters and the second component of seasonal succession parameters Through accumulated years Calculations show that , Years accumulated The value is an integer from 1 to 365, representing the sequential position of the data collection day within the year. The output of the plant health index model is a predicted value of the seasonal physiological response signal of the plant community. .

[0036] In practical implementation, the predicted output of the plant health index model is represented by the following formula:

[0037]

[0038] in, This represents the predicted value of seasonal physiological response signals in plant communities. Indicates the learning rate. The value is 0.05. This represents the total number of regression trees in the model. The value is 150. Indicates the first Regression trees for input feature vectors The output value, This includes estimates of the photosynthetic electron transport rate in the leaves. Air saturated water vapor pressure difference Net radiative flux First component of seasonal succession parameters and the second component of seasonal succession parameters The input feature vector. Tree of return The output value is the value of the regression tree composed of the following elements: The weight value corresponding to the leaf node to which the node falls.

[0039] In practice, the specific steps for training a plant health index model using the extreme gradient boosting regression algorithm are as follows: collect historical data covering the entire annual cycle, construct a training dataset, and each sample in the training dataset contains an estimated value of the photosynthetic electron transport rate of leaves collected synchronously. Air saturated water vapor pressure difference Net radiative flux First component of seasonal succession parameters The second component of seasonal succession parameters and the true values ​​of the corresponding seasonal physiological response signals of plant communities. The true values ​​of the seasonal physiological response signals of the plant community were obtained by weighted fusion of the net photosynthetic rate and transpiration rate of the community measured in the field during the same period, with weighting coefficients set to 0.6 and 0.4, respectively. An additive iterative strategy was used during the training process, with the initial predicted value set to zero. In the [missing information]... In the first iteration, the negative gradient of the residual between the current model's predicted value and the true value is calculated, and this negative gradient is used as the fitting target for the new regression tree; the fitting process continues until the second iteration is complete. When constructing a regression tree, a greedy algorithm is used to split nodes layer by layer, selecting the features and split points that maximize the gain. The gain calculation formula includes first-order gradient statistics and second-order gradient statistics, and an L2 regularization term is added. The regularization parameter... Set to 1.0; the maximum depth of each tree is limited to 5, the minimum sample weight of leaf nodes is set to 1.0, and the column sampling ratio is set to 0.8. Complete. After constructing the regression trees, a trained plant health index model is obtained. To avoid overfitting, an early stopping strategy is adopted during training: the addition of trees is terminated when the error on the validation set no longer decreases after 10 consecutive iterations.

[0040] In practice, the estimated values ​​of leaf photosynthetic electron transport rate, air saturation vapor pressure difference, net radiation flux, and the first and second components of seasonal succession parameters converted from the current date are input into the trained plant health index model to obtain the seasonal physiological response signal of the plant community. This signal is a one-dimensional continuous value, and its level reflects the comprehensive response intensity of the plant community to photosynthetically active radiation interception and water use under the current environmental conditions and seasonal background.

[0041] Example 3:

[0042] In specific implementation, please refer to Figure 4 The seasonal physiological response signal is detrended and periodically decomposed. The seasonal physiological response signal is a one-dimensional time series. ,in The sampling point numbers are arranged in chronological order. The range of values ​​is from arrive , This represents the total length of the seasonal physiological response signal time series. Detrending and periodic decomposition employs an iterative decomposition method based on locally weighted regression. First, the trend component is extracted from the seasonal physiological response signal series. The window width of the locally weighted regression used for trend component extraction is... , The value is set to This corresponds to the yearly cycle length at a daily sampling frequency, allowing the trend component to filter out cyclical fluctuations within a year. Seasonal physiological response signal sequence. Subtracting the extracted trend component yields the detrended seasonal physiological response signal sequence. Periodic subsequence smoothing is then applied to the detrended seasonal physiological response signal sequence to extract the periodic component. The window width of the locally weighted regression used in the periodic subsequence smoothing is [value missing]. , The value is set to The periodic component is extracted. The periodic component is subtracted from the original seasonal physiological response signal sequence to obtain the seasonal physiological response signal sequence after eliminating periodic fluctuations. Then, a trend extraction is performed on the seasonal physiological response signal sequence after eliminating periodic fluctuations to obtain the baseline component. Subtract the baseline component from the original seasonal physiological response signal sequence. The wave components are obtained. The seasonal physiological response signal sequence, baseline component, and fluctuation component satisfy the following relationship:

[0043]

[0044] in, Indicates the first Seasonal physiological response signal values ​​at each sampling point Indicates the first The baseline component values ​​at each sampling point reflect the non-periodic continuous variation trend of the seasonal physiological response signal on interannual and longer scales. Indicates the first The fluctuation component values ​​at each sampling point include the periodic seasonal fluctuations and short-term irregular disturbances in the seasonal physiological response signal.

[0045] In practical implementation, the abrupt change characteristic points of the extract's phenomenological transition nodes are identified. In the fluctuation component... On the sequence, calculate the fluctuation component along the sampling point number. First-order forward difference , Traverse all from arrive The position, when Negative values ​​and When the value is positive, it is determined that the sampling point number is... There exists a local minimum inflection point where the fluctuation component changes from decreasing to increasing. The sampling point number is... A mutation feature point is recorded, and the time position corresponding to the mutation feature point is denoted as . superscript This represents the sequence number of the mutation feature point. For each IoT sensor node, the above mutation feature point extraction process is independently performed on the seasonal physiological response signal sequence calculated from all IoT sensor nodes to obtain the mutation feature point set for each IoT sensor node.

[0046] In practical implementation, based on the temporal localization of multiple abrupt change feature points, the seasonal physiological response signals of all sensor nodes within the garden area are spatially aggregated and temporally aligned. The seasonal physiological response signal sequences of all IoT sensor nodes within the garden area and their corresponding sets of abrupt change feature points are aggregated to the edge computing gateway. The earliest occurrence time point among all abrupt change feature points within the garden area is selected as the time reference point. For each IoT sensor node, the time position of the node's first mutation feature point is compared with the time reference point. The time difference between nodes is used as the offset correction for that node. This offset correction is used to shift the seasonal physiological response signal sequence of the node along the time axis, ensuring that the seasonal physiological response signal sequences of all IoT sensor nodes are time-aligned at the first phenological transition node. At each aligned sampling time, the arithmetic mean of the seasonal physiological response signal values ​​of all IoT sensor nodes is calculated to obtain the population average seasonal physiological response signal value. These population average seasonal physiological response signal values ​​are arranged in chronological order to form the temporal phenological continuum gradient of the park's plant community. The temporal phenological continuum gradient of the park's plant community is a one-dimensional time series, describing the continuous evolution trajectory of the phenological state of the entire park area's plant community over time.

[0047] Example 4:

[0048] In specific implementation, please refer to Figure 5 The gradient of the temporal phenology continuum of the park's plant communities is mapped to a four-stage phenological state space comprising dormancy, budding, vigorous growth, and senescence, thus identifying the current phenological stage of the park. The gradient of the temporal phenology continuum of the park's plant communities is a one-dimensional time series, denoted as... ,in These are the sampling time numbers arranged in chronological order. The range of values ​​is from arrive , This represents the sequence length of the temporal phenological continuum gradient of the plant community in the park. The four phenological stages in the four-stage phenological state space are respectively represented by state variables. The four discrete values ​​represent: Corresponding to the dormancy period, Corresponding to the budding stage, Corresponding to the vigorous growth period, Corresponding to the senescence period. The temporal phenological continuum gradient of the park's plant communities. Mapped to phenological stages The method is as follows: three numerical thresholds are preset, namely the first dividing threshold. Second threshold and the third dividing threshold ,satisfy First threshold The value is the lower quartile of the gradient value interval corresponding to the transition from dormancy to budding in the historical gradient data of the plant community time-series phenology continuum in the park, and the second boundary threshold. The value is taken as the lower quartile of the gradient value interval corresponding to the transition from the budding stage to the vigorous growth stage, and the third boundary threshold. The value is taken as the lower quartile of the gradient value interval corresponding to the transition from the vigorous growth period to the senescence period. All three boundary thresholds were obtained through statistical analysis of historical temporal phenological continuum gradient data of the garden area over the past three years. During real-time mapping, the mean value of the temporal phenological continuum gradient of the plant community within the current sliding time window is calculated. The width of the sliding time window is set to At each sampling time, if Each sampling time corresponds to If the time window is set to one day, then the width of the sliding time window corresponds to a consecutive week. At that time, the current phenological stage of the park is determined to be a dormant period; when At that time, the current phenological stage of the park is determined to be the budding stage; when At that time, it was determined that the current phenological stage of the park belonged to the vigorous growth period; when At that time, the current phenological stage of the park was determined to be the senescence stage. This is how the current phenological stage classification of the park was identified. , Values , , or .

[0049] In practice, it is based on the phenological stage classification. The system retrieves a pre-defined horticultural management rule base and generates a combination of horticultural management actions, including irrigation start / stop thresholds, irrigation depth, pruning frequency, and pruning intensity. The pre-defined horticultural management rule base is stored in the edge computing gateway's storage unit as a structured data table. Each row of the table corresponds to a phenological stage, and the column fields include the phenological stage identifier, initial values ​​for irrigation start / stop thresholds, initial values ​​for irrigation depth, initial values ​​for pruning frequency, and initial values ​​for pruning intensity. During the dormancy period, the corresponding rule records are retrieved, and the initial value of the irrigation start / stop threshold is set when the soil matrix water potential signal is lower than the dormancy period. Start irrigation, higher than Stop irrigation; initial irrigation depth set to [value]. The initial pruning frequency is set to monthly. No pruning is performed at this time, and the initial pruning intensity is set to [value]. That is, do not trim. When During the germination period, the corresponding rule records are retrieved, and the initial value of the irrigation start / stop threshold is set when the soil matrix water potential signal is lower than the threshold value. Start irrigation, higher than Stop irrigation; initial irrigation depth set to [value]. The initial pruning frequency is set to monthly. The initial pruning intensity was set to remove the total canopy biomass. .when During the vigorous growth period, the rule records corresponding to the vigorous growth period are retrieved, and the initial value of the irrigation start / stop threshold is set when the soil matrix water potential signal is lower than the threshold value. Start irrigation, higher than Stop irrigation; initial irrigation depth set to [value]. The initial pruning frequency is set to monthly. The initial pruning intensity was set to remove the total canopy biomass. .when During the aging period, the corresponding rule records are retrieved, and the initial value of the irrigation start / stop threshold is set when the soil matrix water potential signal is lower than the aging period. Start irrigation, higher than Stop irrigation; initial irrigation depth set to [value]. The initial pruning frequency is set to monthly. The initial pruning intensity was set to remove the total canopy biomass. .

[0050] In practice, the soil moisture consumption rate for five consecutive days is extracted from micrometeorological environmental data. Combined with the slope changes of the gradient of the time-series phenological continuum of the park's plant population, the irrigation start / stop threshold is dynamically updated. The extraction method for the five-day soil moisture consumption rate is as follows: Soil matrix water potential signal sequences for the past five days are obtained from the multi-modal heterogeneous data acquisition module. A representative value is taken for each day, and the representative value is the median of the soil matrix water potential signals at all sampling times on that day, thus obtaining the five-day soil matrix water potential value sequence. Soil moisture consumption rate Calculated using the following formula:

[0051]

[0052] in, This represents the soil moisture consumption rate over five consecutive days, in units of... , Indicates five consecutive days of the first The representative value of soil matrix water potential on a given day. The range of values ​​is arrive , Indicates five consecutive days of the first The representative value of soil matrix water potential on a given day. This represents the absolute value of the difference between representative values ​​of soil matrix water potential on two consecutive days. The slope change of the temporal phenological continuum gradient of the plant community in the park is obtained by calculating the slope of the linear fit over seven sampling times prior to the current time, and the absolute value of this slope is denoted as... The dynamic update method for the irrigation start / stop threshold is as follows: the initial value of the irrigation start / stop threshold retrieved from the horticulture management rule base is recorded as the initial value of the start threshold. and initial value of the stopping threshold Updated startup threshold Calculated by the following formula: Updated stopping threshold Calculated by the following formula: .in, This is a correction factor for the soil moisture consumption rate, with a value of [value missing]. The basis for this is that when soil moisture consumption accelerates, irrigation needs to be started earlier and the stop threshold should be increased accordingly to increase the amount of water per irrigation. This is the phenological gradient slope correction coefficient, with a value of [value missing]. The rationale is that when the gradient of the temporal phenological continuum of the plant community in the park rises rapidly, it indicates that the plants are undergoing a rapid phenological transition, requiring synchronized and earlier irrigation to trigger the process. The dynamically updated trigger threshold... and stopping threshold Replace the corresponding initial value in the horticulture management rule base as the actual irrigation start / stop threshold for the current phenological stage.

[0053] In practice, based on the updated irrigation start / stop thresholds, a fuzzy logic controller is invoked to adjust the opening of the irrigation valve, ensuring that the irrigation depth precisely matches the root water absorption layer thickness at the current phenological stage. The fuzzy logic controller employs a dual-input, single-output structure. The first input variable is the soil matrix water potential deviation. Soil matrix water potential deviation Defined as the current soil matrix water potential signal value and the updated trigger threshold The difference, the second input variable is the rate of change of soil matrix water potential deviation. Soil matrix water potential deviation rate Defined as the difference between the soil matrix water potential deviation at the current sampling time and the soil matrix water potential deviation at the previous sampling time. The output variable is the valve opening increment. Valve opening increment The range of values ​​is This indicates a percentage increase or decrease based on the current valve opening. Soil matrix water potential deviation. The fuzzy universe of discourse is divided into five fuzzy sets: negative large, negative small, zero, positive small, and positive large; the rate of change of soil matrix water potential deviation. The fuzzy universe of discourse is divided into five fuzzy sets: negative large, negative small, zero, positive small, and positive large; valve opening increment The fuzzy universe of discourse is divided into five fuzzy sets: negative large, negative small, zero, positive small, and positive large. The fuzzy membership functions all adopt a combination of triangular and trapezoidal functions. The node parameters of the triangular and trapezoidal functions are determined based on the historical fluctuation range of the soil matrix water potential signal and the adjustment accuracy of the irrigation valve opening. The fuzzy rule base contains... There are 10 rules, covering all permutations and combinations of the input fuzzy sets. Each rule is in the form of "If..." yes and yes ,but yes ",in The first term representing the deviation of soil matrix water potential A fuzzy set, The first digit representing the rate of change of soil matrix water potential deviation A fuzzy set, This represents the fuzzy set of valve opening increments activated under corresponding conditions. Fuzzy inference is performed using the Mamdani inference method, and defuzzification is performed using the centroid method to obtain the precise value of the valve opening increment. The edge computing gateway sends the valve opening increment command to the execution controller of the intelligent irrigation valve. The execution controller adds the current valve opening to the received valve opening increment to obtain the target valve opening, and drives the valve actuator to adjust to the target valve opening. Changes in valve opening directly change the irrigation flow rate, thereby controlling the amount of irrigation water per unit time. This ensures that the total amount of water actually infiltrating into the root zone matches the root water absorption layer thickness at the current phenological stage, achieving precise control of irrigation depth.

[0054] Example 5:

[0055] In practice, irrigation and pruning commands are issued to the intelligent irrigation valves and the automated pruning robot arm. The edge computing gateway sends a data packet containing updated irrigation start / stop thresholds, valve opening increments output by the fuzzy logic controller, pruning frequency, and pruning intensity to the programmable logic controllers (PLCs) of the intelligent irrigation valves and the automated pruning robot arm via an RS-485 industrial bus. The PLC of the intelligent irrigation valves parses the data packet and drives a stepper motor to adjust the valve opening; the PLC of the automated pruning robot arm parses the data packet, loads joint motion parameters and end-effector operation parameters, and drives the automated pruning robot arm to perform differentiated horticultural operations in real time.

[0056] In practice, the working path of the automated pruning robot is dynamically planned. The first-order backward difference at the current moment is calculated from the time series of the temporal phenological continuum gradient of the plant community in the park to obtain an estimate of the branch growth rate. Estimated branch growth rate Characterizes the elongation of plant branches per unit time. Estimated based on branch growth rate. The cutting angle of the cutting edge is adjusted in real time. The cutting angle is expressed as the angle between the cutting edge plane and the branch axis. The adjustment relationship of the cutting angle is determined by the following formula:

[0057]

[0058] in, Indicates the angle of entry of the cutting edge of the tool, in degrees. This represents the minimum cutting angle, with values ​​ranging from 1 to 2. , This represents the maximum cutting angle, with values ​​ranging from [value 1] to [value 2]. , This represents the estimated branch growth rate derived from the temporal phenological continuum gradient of the plant community in the park, expressed in millimeters per day. This represents the upper limit of the reference range for branch growth rate, with a value of [value missing]. The value is measured in millimeters per day, determined based on the maximum measured daily elongation of the dominant plant species in the garden area during their vigorous growth period. The cutting distance is estimated based on the branch growth rate. Adjusted by multiplying with the pruning intensity parameter, when Greater than When the cutting stroke is increased by millimeters per day, the cutting stroke is increased from the base value. The automated pruning robot arm's operation path planning employs Dijkstra's algorithm based on grid maps. The 3D point cloud of the canopy in the area to be pruned is projected onto a horizontal grid, with each grid marked as either an obstacle or passable. Taking the current position of the automated pruning robot arm as the starting node and the grid covering all branches to be pruned as the target node set, the algorithm searches for the grid path sequence with the minimum total motion cost. This grid path sequence is then transformed into a continuous and smooth trajectory in the robot arm's joint space, serving as the automated pruning robot arm's operation path.

[0059] In practical implementation, the intelligent irrigation valve is controlled by variable frequency pulses. The actuator of the intelligent irrigation valve receives the valve opening increment from the edge computing gateway and determines the target valve opening. During irrigation, the actuator synchronously receives stem micro-change signals from plant physiological and ecological signals and monitors these signals in real time. When a continuous decrease in the stem micro-change signal is detected within the last thirty minutes, and the decrease exceeds a certain threshold... When the pulse frequency reaches a certain value (in millimeters), the controller switches the output mode of the intelligent irrigation valve from continuous flow to pulsed flow, with the pulse frequency set to a specific value. Hertz, that is, every A pulse is activated every second; when the micro-change signal in the stem rises and the rise amplitude exceeds At millimeters, the pulse frequency decreases to Hertz. In pulsed flow mode, the controller synchronously adjusts the coverage radius of multiple smart sprinkler heads. Each smart sprinkler head has a built-in electromagnetic regulating ring; changing the current in the ring alters the water jet angle. When the pulse frequency is... At a frequency of Hertz, the coverage radius of the smart sprinkler head is adjusted to the nominal radius. When the pulse frequency is At a certain frequency (Hz), the coverage radius of the smart sprinkler head returns to its nominal radius. The flow rate is determined by the opening degree of the smart irrigation valve and the pulse duty cycle, enabling real-time adjustment of the coverage radius and flow rate during irrigation.

[0060] In practical implementation, an IoT-based smart garden management system is provided to execute the aforementioned smart garden management method. A multi-modal heterogeneous data acquisition module is configured on each IoT sensor node and edge computing gateway. It collects multi-modal heterogeneous data in real time through chlorophyll fluorescence sensors, stem micro-change sensors, soil matrix water potential sensors, canopy temperature sensors, and air temperature, humidity, and wind speed sensors. After missing value imputation, abnormal mutation point removal, and standardization, a standardized multi-source time-series data matrix is ​​generated. A seasonal physiological response calculation module extracts estimated leaf photosynthetic electron transport rate, air saturated vapor pressure difference, and net radiation flux from the standardized multi-source time-series data matrix. It then calls a trained plant health index model for nonlinear coupling calculations, outputting the seasonal physiological response signal of the plant community. A phenological continuum construction module receives the seasonal physiological response signal sequences from each IoT sensor node, performs detrending and periodic decomposition, extracts mutation feature points, and spatially aggregates and temporally aligns the seasonal physiological response signals of all sensor nodes to generate a temporal phenological continuum gradient for the park's plant community. The management mode switching module maps the gradient of the plant community's temporal phenological continuum to a four-stage phenological state space including dormancy, budding, vigorous growth, and senescence. It identifies the current phenological stage of the park, retrieves a preset horticultural management rule base, dynamically updates irrigation start / stop thresholds, and calls a fuzzy logic controller to generate valve opening increments, forming a combination of horticultural management actions including irrigation start / stop thresholds, irrigation depth, pruning frequency, and pruning intensity. The intelligent device drive module encodes the horticultural management action combination into a work instruction data packet, which is transmitted to the intelligent irrigation valve and automatic pruning robot arm via an industrial bus. This drives the intelligent irrigation valve to perform variable frequency pulse irrigation and the automatic pruning robot arm to perform differentiated pruning operations. The closed-loop feedback calibration module receives the actual cumulative flow data from the intelligent irrigation valve and the actual joint movement trajectory and pruning area image data from the automatic pruning robot arm via IoT communication. The closed-loop feedback calibration module calculates the deviation between the actual cumulative flow data and the water demand estimated based on plant physiological and ecological signals. When the deviation exceeds... At that time, a valve opening correction coefficient is generated and sent to the intelligent device drive module, so that the valve opening increment in the next round of irrigation operation instructions is scaled according to the correction coefficient. The closed-loop feedback calibration module performs canopy coverage analysis on the pruning area image, compares the actual canopy coverage obtained from the analysis with the target canopy coverage, and when the difference exceeds the preset tolerance, it sends a pruning intensity adjustment prompt to the management mode switching module, triggering the correction of the pruning intensity parameter in subsequent pruning operation instructions.

[0061] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A smart garden management method based on the Internet of Things, characterized in that, Includes the following steps: Based on multiple IoT sensor nodes deployed within the garden area, real-time collection of plant physiological and ecological signals and micro-meteorological environment data is achieved. The photosynthetically active radiation interception efficiency of the plant physiological and ecological signals is analyzed, and the seasonal physiological response signals of the plant community are calculated by combining the evapotranspiration parameters in the micrometeorological environment data. Specifically, this includes: The estimated value of leaf photosynthetic electron transfer rate is extracted from the plant physiological and ecological signals, and the air saturated water vapor pressure difference and net radiation flux are extracted from the micrometeorological environment data to construct a plant health index model that integrates the prior knowledge of seasonal evolution. The plant health index model is invoked to perform nonlinear coupling calculations on the estimated value of the photosynthetic electron transport rate of the leaves, the air saturation vapor pressure difference, and the net radiation flux, and outputs the seasonal physiological response signal of the plant community. Based on the aforementioned seasonal physiological response signals, a temporal phenological continuum gradient characterizing plant phenological evolution is constructed, specifically including: The seasonal physiological response signal is detrended and decomposed into periods to separate the baseline and fluctuation components and extract abrupt change feature points of the phenological transition nodes. Based on the temporal localization of multiple mutation feature points, the seasonal physiological response signals of all sensor nodes in the garden area are spatially aggregated and temporally aligned to generate a temporal phenological continuum gradient of the garden plant community. Based on the gradient of the temporal phenological continuum, the horticultural management mode is dynamically switched to generate irrigation and pruning operation instructions that match the current phenological stage; The irrigation and pruning operation commands are issued to the intelligent irrigation valves and automatic pruning robotic arms, driving each intelligent device to perform differentiated horticultural operations in real time, specifically including: The working path of the automatic pruning robot arm is dynamically planned, and the cutting angle and cutting stroke of the cutting edge are adjusted in real time according to the branch growth rate described in the temporal phenological continuum gradient. The intelligent irrigation valve is controlled by a variable frequency pulse. Based on the stem micro-change signal in the plant physiological and ecological signal, the coverage radius and flow rate of multiple intelligent sprinkler heads are adjusted in real time during the irrigation process.

2. The smart garden management method based on the Internet of Things according to claim 1, characterized in that, The method, based on multiple IoT sensor nodes deployed within the garden area, collects plant physiological and ecological signals and micro-meteorological environment data in real time, specifically including: Multimodal heterogeneous data is collected for each of the IoT sensor nodes. The multimodal heterogeneous data includes chlorophyll fluorescence-induced dynamics signals, stem micro-change signals, soil matrix water potential signals, canopy temperature signals, and air temperature, humidity, and wind speed signals. The collected multi-modal heterogeneous data is imputed for missing values ​​and anomalous mutation points are removed to generate a standardized multi-source time series data matrix.

3. The smart garden management method based on the Internet of Things according to claim 1, characterized in that, The method of dynamically switching horticultural management modes based on the temporal phenological continuum gradient and generating irrigation and pruning operation instructions matching the current phenological stage specifically includes: The temporal phenological continuum is gradient-mapped to a four-stage phenological state space comprising dormancy, budding, vigorous growth, and senescence, thereby identifying the current phenological stage affiliation of the park. Based on the phenological stage classification, a preset horticultural management rule library is retrieved to generate a combination of horticultural management actions, including irrigation start / stop thresholds, irrigation depth, pruning frequency, and pruning intensity.

4. The smart garden management method based on the Internet of Things according to claim 3, characterized in that, Based on the phenological stage classification, a preset horticultural management rule base is retrieved to generate a combination of horticultural management actions, including irrigation start / stop thresholds, irrigation depth, pruning frequency, and pruning intensity. Specifically, this includes: The soil moisture consumption rate for five consecutive days is extracted from the micrometeorological environment data, and the irrigation start / stop threshold is dynamically updated by combining the slope change of the gradient of the time-series phenological continuum. Based on the updated irrigation start / stop threshold, the fuzzy logic controller is invoked to adjust the opening of the irrigation valve, so that the irrigation depth accurately matches the root water absorption layer thickness at the current phenological stage.

5. The smart garden management method based on the Internet of Things according to claim 1, characterized in that, The plant health index model uses an extreme gradient boosting regression algorithm based on seasonal succession parameters to predict the health index with the estimated leaf photosynthetic electron transport rate, the air saturated vapor pressure difference, and the net radiation flux as input variables.

6. A smart garden management system based on the Internet of Things, characterized in that, A smart garden management method based on the Internet of Things (IoT) as described in any one of claims 1-5, comprising: The multi-mode heterogeneous data acquisition module is used to collect plant physiological and ecological signals and micro-meteorological environment data in real time through multiple IoT sensor nodes deployed in the garden area; The seasonal physiological response calculation module is used to analyze the photosynthetically active radiation interception efficiency of the plant physiological and ecological signals, and calculate the seasonal physiological response signals of the plant community by combining the evapotranspiration parameters in the micrometeorological environment data. The phenological continuum construction module is used to construct a temporal phenological continuum gradient characterizing the evolution of plant phenology based on the seasonal physiological response signals. The management mode switching module is used to dynamically switch the horticultural management mode according to the gradient of the time-series phenological continuum, and generate irrigation and pruning operation instructions that match the current phenological stage. The intelligent device driving module is used to send the irrigation and pruning operation instructions to the intelligent irrigation valve and the automatic pruning robotic arm, driving them to perform differentiated horticultural operations. The closed-loop feedback calibration module is used to receive the operation feedback data transmitted back from the intelligent device, cross-validate it with the plant physiological and ecological signals, and trigger the correction of subsequent operation instructions.

Images