Garden intelligent management system and method based on Internet of Things

Abstract

The invention relates to the technical field of intelligent gardens and ecological monitoring, and particularly discloses an intelligent garden management system and method based on the Internet of Things, and the method comprises the steps: carrying out the correlation analysis of environment parameter values monitored by all sensor nodes in a historical period of a garden and microtopography factors; whether the problem of insufficient monitoring data representativeness caused by microtopography heterogeneity exists or not is judged, if yes, grid units are divided, the grid units which have similar environment characteristics and are adjacent in space are divided into one microhabitat, and master and slave monitoring nodes are arranged for each microhabitat; according to an environment response mode, monitoring nodes are clustered into a plurality of response clusters, a management unit is divided, a microhabitat prediction model driven by a master monitoring node is constructed, data of slave monitoring nodes are predicted according to data of the master monitoring node, and the problem of insufficient representativeness of single-point monitoring data caused by garden microtopography heterogeneity is solved. And low-cost and high-precision comprehensive perception and differential precise management of the garden environment are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of smart garden and ecological monitoring technology, and particularly relates to a smart garden management system and method based on Internet of Things. BACKGROUND

[0002] With the development of Internet of Things technology, the intelligent management of gardens has gradually become popular. By deploying various types of environmental sensors (such as soil moisture, temperature, and light sensors) in the garden, managers can remotely and in real time understand the environmental status of the garden and make decisions such as irrigation and fertilization accordingly.

[0003] However, in practical applications, the existing technology faces a common and difficult problem: the lack of representativeness of monitoring data caused by the microtopographic heterogeneity of the garden. Gardens are not homogeneous spaces, and due to factors such as topographic relief, vegetation cover, and water distribution, they naturally form numerous microhabitats with clear boundaries. For example, sunny slopes and shady depressions, tree canopy and open lawns, water edge and dry areas, their soil moisture, temperature, and light parameters often differ significantly (local differences can exceed 30%). The existing technology usually deploys sensors using uniform or empirical distribution methods, resulting in data collected by a single sensor only reflecting the environmental conditions in a very small area, and not representing the surrounding area, thus leading to misjudgments in management decisions. For example, irrigating the entire garden based on the dry data of the open lawn may cause waterlogging and root rot in the already moist forest area.

[0004] To solve the problem of data representativeness, one intuitive method is to increase the density of sensor deployment to achieve comprehensive monitoring. However, this will result in a sharp increase in hardware costs, installation and maintenance costs, and data transmission and processing costs, which is difficult for large-area gardens to bear.

[0005] Therefore, the present application provides a smart garden management system and method based on Internet of Things. SUMMARY

[0006] The present application aims to provide a smart garden management system and method based on Internet of Things to solve the above-mentioned background problems.

[0007] The purpose of the present application can be achieved by the following technical solution: a smart garden management method based on Internet of Things, comprising the following steps: correlating and analyzing the environmental parameter values collected by multiple sensor nodes in the garden and the microtopographic factors in the historical period to determine whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity; if so, dividing the garden space into multiple microhabitats according to the microtopographic factors and environmental characteristics; The monitoring nodes are deployed in the area selected by the representative analysis of the micro-terrain factors and the environmental characteristics in the micro-habitat, a master monitoring node and multiple slave monitoring nodes are set for each micro-habitat, and the garden is monitored; The monitoring data of each master and slave monitoring node are collected and analyzed, the environmental response modes of each master and slave monitoring node are extracted, and each monitoring node is clustered into multiple environmental response mode groups according to the environmental response modes; The monitoring nodes and the corresponding micro-habitats belonging to the same environmental response mode group are integrated into a management unit, and a micro-habitat prediction model driven by the master monitoring node is constructed to support differentiated maintenance decisions.

[0008] Further, the process of the correlation analysis is: The management data in the historical period of the garden are acquired, including: Micro-terrain factors: elevation, slope, aspect, surface curvature, surface roughness; Environmental parameters monitored by each sensor node: soil moisture, soil temperature, air temperature and humidity, and light intensity, which need to include the spatial position information of each node; The coefficient of variation of the environmental parameters within a 5m range centered on the sensor node is calculated, and the sensor nodes whose coefficient of variation meet the requirements are marked as discrete nodes; A distance weight matrix is constructed according to the coordinates of each sensor node, and the Moran index is calculated according to the Moran index calculation formula, if the Moran index is less than or equal to 0, the garden has micro-terrain heterogeneity; For any sensor node: The Pearson correlation coefficient of the environmental parameter variation coefficient and the micro-terrain factor value is calculated, if the Pearson correlation coefficient meets the requirements, it is a strong correlation.

[0009] Further, the judgment method of whether there is a problem of insufficient representativeness of monitoring data caused by micro-terrain heterogeneity is: The proportion of discrete nodes in all sensor nodes is calculated to obtain the proportion of discrete nodes; If the proportion of discrete nodes meets the requirements, the garden has micro-terrain heterogeneity, and at least one micro-terrain factor has a strong correlation with the environmental parameter variation coefficient, then there is a problem of insufficient representativeness of monitoring data caused by micro-terrain heterogeneity.

[0010] Further, the way of dividing the garden space into multiple micro-habitats is: The garden is divided into multiple grid units, the micro-terrain factors and environmental parameters of each grid unit are standardized, the grid units that are adjacent in space are first classified into candidate groups, then the feature similarity between the grid units in the candidate groups is calculated for feature clustering, and finally the clustering unit is the micro-habitat.

[0011] Further, the selection method of the monitoring node deployment area is: Calculate the microtopographic factor mean value and the environmental parameter mean value of each microhabitat, conduct field reconnaissance on the microhabitat whole domain, exclude the areas where the microtopographic factor exceeds the microhabitat mean value and the environmental parameter exceeds the microhabitat environmental parameter mean value, and the remaining area is the candidate area; Uniformly arrange 3 measurement points in each candidate area, measure the microtopographic factor mean value and the environmental parameter, and calculate the average value to obtain the candidate area mean value; Calculate the relative deviation of the candidate area mean value and the microhabitat mean value, if the relative deviation of all microtopographic factors and environmental parameters meets the requirements, it is the main monitoring node deployment area.

[0012] Further, the selection method of the monitoring node deployment area is: Calculate the microtopographic factor mean value of the microhabitat as the microtopographic reference value, calculate the deviation of the microtopographic factor of each grid unit in the microhabitat from the microtopographic reference value, and mark the grid unit with a deviation meeting the requirements as a microtopographic difference area; Calculate the environmental parameter mean value of the microhabitat as the environmental reference value, measure the environmental parameter in the microtopographic difference area, calculate the deviation of the environmental parameter of each grid unit in the microtopographic difference area from the environmental parameter reference value, and mark the grid unit with a deviation meeting the requirements as an environmental difference area; In each environmental difference area, arrange multiple measurement points to measure the microtopographic factor and the environmental parameter; Calculate the parameter mean value of the measurement points, select the point closest to the area mean value as the slave monitoring node, and the straight line distance between the slave monitoring node and the main monitoring node does not exceed the maximum radius of the microhabitat.

[0013] Further, the extraction method of the environmental response mode of each main and slave monitoring node is: From the time series monitoring data of each main and slave monitoring node, extract the daily variation feature, seasonal change trend feature, extreme weather response feature and external correlation feature, specifically including: Daily variation feature: diurnal fluctuation amplitude of parameter, peak value occurrence time, valley value occurrence time; Seasonal change trend feature: seasonal rising rate, seasonal falling rate, seasonal mean value of parameter; Extreme weather response feature: response amplitude = | extreme weather parameter extreme value - extreme weather parameter reference value |, response rate = response amplitude / time interval from extreme weather to parameter reaching extreme value; External correlation feature: Pearson correlation coefficient of node parameter and external climate factor.

[0014] Further, the method of clustering each monitoring node into multiple environmental response mode groups is: The four types of environmental response characteristics, i.e., daily variation characteristics, seasonal variation trend characteristics, extreme weather response characteristics and external correlation characteristics, are standardized and used as clustering characteristics, and clustering is performed, and finally a plurality of environmental response mode groups are obtained.

[0015] Further, the main monitoring node driven microhabitat prediction model is constructed and predicted in the following manner: The main monitoring node, the slave monitoring node and the microhabitat to which they belong in the same environmental response mode group are integrated into a management unit. The monitoring data of the main monitoring node and the slave monitoring node are integrated into a feature set, which specifically includes: Core driving features: real-time values, change rates and cumulative values of key parameters of the main monitoring node; Spatial correlation features: straight-line distance, elevation difference, slope difference and whether belonging to the same response cluster between the slave monitoring node and the main monitoring node; Environmental background features: soil type, type of ecological unit to which it belongs, annual average sunshine duration, soil organic matter content and surface coverage type; The monitoring data are aligned according to the time stamp, and the corresponding feature set is matched for each slave monitoring node to form a complete training data set; The random forest is used to train the model using the training data set, and a main monitoring node driven microhabitat prediction model is obtained. The main monitoring node collects real-time data to generate core driving features, and retrieves spatial correlation features and environmental background features of the slave monitoring node, and inputs the main monitoring node driven microhabitat prediction model, and the model outputs the predicted value of the key parameter of the slave monitoring node in real time.

[0016] An intelligent garden management system based on the Internet of Things, comprising the following modules: A problem judgment module: correlation analysis is performed on the environmental parameter values collected by a plurality of sensor nodes in the garden and the microtopographic factors in a historical period to determine whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity; A microhabitat division module: if there is, the garden space is divided into a plurality of microhabitats according to the microtopographic factors and environmental characteristics; A monitoring node setting module: the representative analysis of the microtopographic factors and environmental characteristics in the microhabitat is performed to select the monitoring node deployment area, and a main monitoring node and a plurality of slave monitoring nodes are set for each microhabitat to monitor the garden; A response mode group division module: the monitoring data of each main and slave monitoring node are collected and analyzed, the environmental response mode of each main and slave monitoring node is extracted, and each monitoring node is clustered into a plurality of environmental response mode groups according to the environmental response mode; A prediction model construction module: the monitoring nodes belonging to the same environment response mode group and the corresponding microhabitats are integrated into a management unit, and a microhabitat prediction model driven by the main monitoring node is constructed to support differentiated maintenance decisions.

[0017] The beneficial effects of the present application are as follows: By the strategy of microhabitat division-main monitoring node representation, a limited number of main monitoring nodes represent each homogeneous unit, and a small number of slave monitoring nodes capture internal variations, replacing the expensive global dense point layout, significantly reducing costs while ensuring the scientific representativeness of monitoring data for complex garden environments. The introduction of the concept of environment response mode not only focuses on the instantaneous value of the environmental parameter, but also focuses on its dynamic response behavior to events such as rainfall and high temperature, upgrading management from static state to dynamic behavior, thereby providing a basis for adaptive management by more deeply understanding the ecological functions of different regions.

[0018] Through microhabitat division and response mode clustering, the garden is reorganized from geographical space into multiple management units, each unit has consistent needs, achieving precise management of one unit-one strategy, avoiding resource waste and ecological disturbance, and solving the fundamental contradiction between unified management and differentiated needs. Using main monitoring node data to drive prediction models, real-time inference of slave monitoring nodes and regional states that are not directly monitored, upgrading the system from monitoring known to predicting unknown, achieving forward-looking management and covering the blind area of the monitoring network. BRIEF DESCRIPTION OF DRAWINGS The present application will be further described below with reference to the accompanying drawings.

[0019] Figure 1 is a flowchart of a garden intelligent management method based on the Internet of Things according to Embodiment 1 of the present application; Figure 2 is a logic judgment diagram for judging whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity in Embodiment 1 of the present application. Figure 3 is a functional module diagram of a garden intelligent management system based on the Internet of Things according to Embodiment 2 of the present application. DETAILED DESCRIPTION

[0020] In order to make the technical means, creative features, purposes and effects achieved by the present application easy to understand, the present application will be further described below with reference to the specific embodiments.

[0021] Embodiment 1: Please refer to Figure 1 The garden intelligent management method based on the Internet of Things according to the embodiments of the present application specifically includes the following steps: Step one: Correlation analysis of environmental parameter values collected by multiple sensor nodes in the garden during the historical period and micro-topographic factors to determine whether micro-topographic heterogeneity leads to insufficient representativeness of monitoring data; Please refer to Figure 2 As shown in step one, the process of correlation analysis of environmental parameter values and micro-topographic factors includes: Obtain management data during the historical period of the garden, including: Micro-topographic factors: elevation, slope, aspect, surface curvature, surface roughness; Environmental parameters monitored by each sensor node: soil moisture, soil temperature, air temperature and humidity, light intensity, including the spatial location information of each node; First, select the environmental parameters or adjacent node data within a 5m range around each sensor node as samples, calculate the coefficient of variation for each node, and determine the local data dispersion, specifically: For any sensor node: Calculate the coefficient of variation by proportioning the standard deviation of environmental parameters within a 5m range around the sensor node to the arithmetic mean; Compare the coefficient of variation with the preset coefficient of variation. If the coefficient of variation is greater than the preset coefficient of variation, mark the sensor node as a discrete node; It can be understood that the coefficient of variation is a dispersion index that eliminates the influence of dimension. The larger the value, the greater the difference in environmental parameters within a 5m range around the sensor node. In the garden, a coefficient of variation greater than 0.2 indicates high dispersion, and single-node data cannot represent the true environment of the surrounding area; Second, construct a spatial weight matrix based on the coordinates of each sensor node, calculate the Moran's I index for each environmental parameter, and perform statistical tests to determine whether the correlation is significant, specifically: According to the coordinates of each sensor node, calculate the straight-line distance between any two sensor nodes, determine the value of w ij according to the distance threshold, form an n x n weight matrix, where w ij represents the distance weight of the i, j sensor node: if the straight-line distance between the i, j sensor nodes is less than or equal to 5m, w ij = 0, otherwise w ij = 1, and when i = j, w ij = 0, i.e. the sensor node has no weight; Calculate the global Moran's I index I of the garden according to the Moran's I calculation formula , where n is the total number of sensor nodes, x i and x j are the means of the i, j sensor node parameters, w is the mean value of all sensor node parameters ij is a spatial weight matrix; It can be understood that the physical meaning of the Moran index I is that the Moran index is a core index for measuring spatial correlation, the value range is [-1, 1], I>0 indicates spatial positive correlation, adjacent node data are similar (regions with uniform microtopography will exhibit this feature), I<0 indicates spatial positive correlation, adjacent node data are different (regions with complex microtopography will exhibit this feature), and I=0 indicates spatial correlation, and data are randomly distributed (indicating that the global environment heterogeneity is significant); If the Moran index is less than or equal to 0, the garden has microtopographic heterogeneity; Thirdly, it is to be explained that the correlation analysis is performed on the variation coefficient of each sensor node and the microtopographic factor one by one to verify the correlation between the heterogeneity and the microtopography, and specifically: For any sensor node: The Pearson correlation coefficient of the environmental parameter variation coefficient and the microtopographic factor value is calculated, and compared with a preset correlation coefficient, if the Pearson correlation coefficient is greater than the preset correlation coefficient, it is a strong correlation; In step one, the process of judging whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity includes: The proportion of the discrete node in all sensor nodes is counted to obtain the proportion of the discrete node; The proportion of the discrete node is compared with a preset proportion, if the proportion of the discrete node is greater than or equal to the preset proportion, and the variation coefficient has a strong correlation with at least one microtopographic factor, there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity; It should be noted that the function of judging whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity is to accurately distinguish between real data heterogeneity caused by microtopography and false heterogeneity caused by sensor failure / acquisition error, and to avoid deviation of subsequent work; Step two: if there is, the garden space is divided into multiple microhabitats according to the microtopographic factor and the environmental characteristics; In step two, the process of dividing the garden space into multiple microhabitats according to the microtopographic factor and the environmental characteristics includes: The garden is divided into multiple grid units according to the garden area and the microtopographic complexity, the core principle is that the grid can cover the smallest heterogeneous unit of the microtopography, for example: The region with complex microtopography (such as slope and water system periphery) selects a fine grid of 1m×1m, and the region with gentle microtopography (such as flat lawn) selects a coarse grid of 5m×5m, the grid size is a key parameter affecting the partitioning accuracy, too small will increase the calculation amount, and too large will not be able to identify the microtopographic difference; Assign a unique code (e.g., row number + column number) to each grid and record the latitude, longitude, and elevation of the grid center point to ensure spatial traceability of the grid; Extract micro-topographic factors for each grid cell, and use Kriging interpolation (an interpolation method based on spatial autocorrelation, suitable for continuous parameter completion in garden environment) to interpolate historical environmental parameters to each grid to obtain environmental characteristics such as soil moisture, temperature, and humidity of the grid; Standardize all micro-topographic factors and environmental parameters using Z-score to eliminate the effects of dimension and order of magnitude; Calculate the feature similarity between grid cells using Euclidean distance, and use hierarchical clustering to first group spatially adjacent grid cells into candidate groups, and then perform feature clustering within the candidate groups based on a threshold of Euclidean distance < a pre-set distance (after standardization). The final clustered units are microhabitats. It should be noted that the purpose of dividing microhabitats is to refine the garden management unit from the global to the microhabitat, with each microhabitat having uniform environmental and topographic characteristics, providing a spatial foundation for subsequent differentiated management. A feature file of microhabitats is established, converting abstract environmental heterogeneity into specific numerical indicators, and changing garden management from experience-driven to data-driven. Step three: Select monitoring node deployment areas by performing representative analysis of micro-topographic factors and environmental characteristics within the microhabitat, set one main monitoring node and multiple slave monitoring nodes for each microhabitat, and monitor the garden; In step three, the process of setting one main monitoring node and multiple slave monitoring nodes for each microhabitat to monitor the garden includes: First, the selection of the main monitoring node: The main monitoring node is the core representative of the microhabitat, and the selection principle is feature typicality + ecological criticality + topographic representativeness. Each microhabitat is set with only one main monitoring node, and the specific selection criteria are: Calculate the mean of micro-topographic factors and environmental parameters of each microhabitat, and exclude areas where micro-topographic factors exceed the mean of the microhabitat by ±20% and environmental parameters exceed the mean of the microhabitat by ±20%. In the remaining areas, preferentially select areas with complex ecological processes and key plant growth, and the specific judgment criteria are: Plant root zone: within the projection range of tree / shrub canopy, confirmed by vegetation cover (> 70%) and species type (dominant species of the microhabitat); Soil moisture exchange active area: areas with slight dry branches and leaves covering the surface, no standing water but surface moisture (this type of area is the core area of microhabitat water infiltration and evaporation); Arboretum-grassland transition zone: the transition zone between trees and shrubs, and shrubs and lawns within the microhabitat, where the ecological process is complex, and environmental changes can quickly reflect the ecological status of the microhabitat; In the selected ecological key area, 3 measurement points were evenly distributed in each region (to avoid single point error), and core parameters were measured using high-precision equipment. The original data of each measurement point were recorded. For any microtopographic factor: The arithmetic mean of the microtopographic factors of the three measurement points was calculated to obtain the mean value of the candidate area. The relative deviation of the mean value of the candidate area and the microhabitat was calculated. If the relative deviation of all microtopographic factors is less than or equal to the preset deviation, and the relative deviation of all environmental parameters is less than or equal to 10%, the microtopographic factors and environmental parameters are highly matched with the mean value of the microhabitat, i.e., the main monitoring node. The spatial coordinates, microtopographic factor values, and microhabitat codes of the main monitoring node were recorded, and a one-to-one correspondence between the main monitoring node and the microhabitat was established. Second, the deployment of monitoring nodes: The monitoring nodes are used to reflect the environmental heterogeneity within the microhabitat. The deployment principle is heterogeneity coverage + cost control, which includes: Using the mean value of the microhabitat as the baseline value, potential areas with significant differences from the baseline were selected by quantitative threshold: The mean value of the microhabitat's microtopographic factors was calculated as the microtopographic baseline value. The deviation of each grid cell's microtopographic factors from the microtopographic baseline value was calculated. Grid cells with a deviation greater than the microtopographic threshold were marked as microtopographic difference areas. The mean value of the microhabitat's environmental parameters was calculated as the environmental baseline value. The environmental parameters were measured in the microtopographic difference areas. The deviation of each grid cell's environmental parameters from the environmental baseline value was calculated. Grid cells with a deviation greater than the environmental threshold were marked as environmental difference areas. In each environmental difference area, 3-5 measurement points were arranged in a 5m x 5m grid to measure microtopographic factors and environmental parameters. Selecting regional feature representative points: Calculate the parameter mean value of the measurement points, and select the point closest to the regional mean value as the candidate point (to avoid selecting extreme points within the region, such as the deepest water point in a low-lying area). Determine the number based on the area and internal heterogeneity of the microhabitat, for example: set 2-3 from monitoring nodes for microhabitat areas <100㎡, 4-6 for areas 100-500㎡, and 6-8 for areas >500㎡. Deploy environmental feature difference points within the microhabitat, such as high slopes, low-lying areas, and light difference areas. The straight-line distance between the from monitoring nodes and the main monitoring nodes should not exceed the maximum radius of the microhabitat (to ensure the environmental relevance of the main and from monitoring nodes). The master-slave monitoring nodes use the same type and accuracy of sensors to avoid data deviation caused by equipment differences; Thirdly, it is necessary to explain that standardized monitoring is implemented to ensure the timeliness and integrity of data, specifically: The monitoring frequency is set according to the time variability of environmental parameters. The higher the variability, the higher the monitoring frequency. For example: | Parameter Type | Specific Index | Monitoring Frequency | | High-frequency dynamic parameters | Soil moisture, soil temperature, air temperature and humidity | 15 minutes | | Medium-frequency parameters | Light intensity, soil electrical conductivity | 1 hour | | Low-frequency static parameters | Soil pH value, organic matter content | 1 day | The monitoring period covers at least one complete growing season to ensure that the data includes different weather conditions such as rainy season, dry season, high temperature, and low temperature, providing sufficient samples for subsequent response mode analysis; Store according to four-level dimensions of microhabitat code-node type (master / slave)-node number-time stamp, and synchronously record external climate data (rainfall, atmospheric temperature, wind speed) during monitoring as external driving factors for subsequent environmental response mode analysis; It should be noted that the arrangement of master-slave monitoring nodes and monitoring is to replace global dense deployment with a low-cost deployment mode of 1 master + multiple slaves, significantly reducing sensor procurement, installation, and operation and maintenance costs. The master monitoring node ensures accurate monitoring of the core characteristics of the microhabitat, and the slave monitoring node covers internal heterogeneity. The combination of the two makes the data both representative and comprehensive; Step four: Collect and analyze the monitoring data of each master and slave monitoring node, extract the environmental response mode of each master and slave monitoring node, and cluster each monitoring node into multiple environmental response mode groups according to the environmental response mode; In step four, the process of extracting the environmental response mode of each master and slave monitoring node includes: Firstly, the environmental response mode extraction quantifies the dynamic response law of the node, specifically: From the time series monitoring data of each master and slave monitoring node, four types of environmental response characteristics are extracted to form the environmental response mode of the node. All characteristics are around the dynamic response of node parameters to external environment: Daily variation law feature: amplitude of day-night fluctuation (maximum value-minimum value) of parameter, peak value occurrence time, valley value occurrence time; Seasonal variation trend feature: seasonal rising rate of parameter (such as monthly average rising amount of soil moisture in rainy season), seasonal falling rate (such as monthly average falling amount of soil moisture in dry season), and seasonal average; Extreme weather response feature: First, define the extreme weather in the garden (rainfall ≥ 20 mm / 24h is heavy rain, temperature ≥ 35℃ is high temperature, continuous 15 days without rainfall is drought), and then extract two core indicators: Response amplitude = | extreme weather after parameter extreme value - extreme weather before parameter baseline value |, reflecting the sensitivity of node environment to extreme weather, the larger the amplitude, the higher the sensitivity; Response rate = response amplitude / time interval from extreme weather to parameter reaching extreme value, reflecting the speed of node environment response to extreme weather, the faster the rate, the faster the adjustment of environmental parameters; External correlation feature: Pearson correlation coefficient of node parameters and external climate factors, quantifying the correlation between node environmental parameters and global climate, the larger the Pearson correlation coefficient, the more significant the impact of external climate on node environment; In step four, the process of clustering each monitoring node into a plurality of environment response mode groups according to the environment response mode includes: The four types of environment response features, daily variation regularity feature, seasonal variation trend feature, extreme weather response feature, and external correlation feature, are standardized as clustering features, and the specific clustering process includes: Determine the number of clusters K using the elbow method, K starts from 1, calculate the sum of squared errors SSE corresponding to each K, draw the K-SSE curve, and the point where SSE drops sharply and then flattens corresponds to the optimal value of K; Randomly select K clustering features as initial cluster centers, calculate the cosine similarity between each clustering feature and the K cluster centers, and assign the clustering feature to the cluster with the closest cosine similarity; After all clustering features are assigned, recalculate the center of each cluster, repeat the assignment and update until the cluster center change is less than or equal to the preset threshold, and the clustering ends, finally obtaining a plurality of environment response mode groups; For each response cluster, extract common response features and spatial distribution patterns to form a cluster feature profile; It should be noted that the role of clustering according to the environment response mode is to merge a large number of monitoring nodes into a few response clusters, and the subsequent prediction model does not need to model each node individually, but only needs to build a model for each environment response mode group, greatly reducing the computational load and difficulty of model training. The common features of the response cluster reflect the areas in the garden with similar environmental response patterns, providing data support for subsequent cluster management; Step five: integrate monitoring nodes and corresponding microhabitats belonging to the same environment response mode group into a management unit, and build a microhabitat prediction model driven by the main monitoring node to support differentiated maintenance decisions; In step five, the process of integrating monitoring nodes belonging to the same environmental response mode group and their corresponding microhabitats into a single management unit includes: The master and slave monitoring nodes and their respective microhabitats of the same environmental response mode group are integrated into a management unit, and each management unit is assigned a unique code. The characteristic files of all microhabitats within the management unit and the common characteristics of the environmental response mode group are integrated to form a comprehensive characteristic file of the management unit, which clarifies the environmental characteristics, response patterns and core management needs of the unit. Understandably, the basis for dividing management units is that microhabitats with consistent response patterns have consistent environmental change patterns, plant growth needs, and maintenance and management measures. After being integrated into management units, a unified maintenance strategy can be adopted to avoid a one-size-fits-all approach. Based on the comprehensive feature profiles of management units, targeted maintenance strategies are developed, including irrigation timing / volume, fertilizer type / frequency, key points of pest and disease control, and vegetation pruning cycle, to achieve precise management with one strategy per unit. In step five, the construction process of the microhabitat prediction model driven by the main monitoring node includes: The first point to clarify is that defining the forecasting target focuses on core management decision parameters, specifically: Strongly correlated with management decisions: Select parameters that directly influence landscape maintenance decisions; Significantly influenced by micro-topography: Select parameters that reflect the heterogeneity of microhabitats; Appropriate number: Select 2-3 core parameters to avoid excessive targets that could increase model complexity; Prioritize soil moisture content (which determines irrigation timing and amount), surface temperature (which affects plant transpiration and the occurrence of pests and diseases), and relative humidity (which affects the spread of fungal diseases). Prediction granularity determination: Time granularity: Consistent with the monitoring frequency of the main monitoring node to ensure the timeliness of the prediction data; Spatial granularity: Taking the monitoring nodes as the prediction objects, output the parameter prediction values ​​of each monitoring node to ensure the spatial accuracy of the prediction data; Secondly, it should be noted that the construction of the feature set is as follows: The feature set is the core input of the prediction model and needs to fully integrate dynamic monitoring data, spatial relationship data, and static background data. It is divided into three core feature categories. All features need to be preprocessed before being input into the model, specifically including: Core driving features: Real-time values, rate of change, and cumulative values ​​of key parameters of the main monitoring node are the core basis for model prediction and the direct driving factors of parameter changes from the monitoring node; the rate of change reflects the dynamic trend, and the cumulative value reflects the long-term impact. Spatial correlation features: straight-line distance from the monitoring node to the main monitoring node, elevation difference, aspect difference, whether they belong to the same response cluster; Environmental background features: soil type, ecological unit type, average annual sunshine duration, soil organic matter content, land cover type; Thirdly, it is necessary to explain that model training, specifically: Align the monitoring data by timestamp, match the corresponding core driving features, spatial correlation features, environmental background features (input) and key parameter measured values (output) for each slave monitoring node, and form a complete training data set; Divide the data set into training set (used for model parameter learning), validation set (used for hyperparameter tuning), and test set, and ensure that the time distribution of each data set is uniform (such as including rainy season and dry season data in each data set); According to the characteristics of garden environment data (mixed numerical / categorical features, containing time series features), select random forest as the model; Input the training set into the model, with the goal of minimizing the deviation between the predicted value and the measured value, iteratively update the model parameters until the performance of the model on the validation set tends to be stable; Fourthly, the prediction process is: The main monitoring node collects data in real time to generate core driving features (real-time value, change rate, cumulative value); Retrieve the spatial correlation features (distance, elevation difference, whether they belong to the same cluster) and environmental background features (soil type, ecological unit, etc.) of the slave monitoring node from the pre-stored database; Input the three types of features into the trained prediction model, and the model outputs the predicted value of the key parameters of the slave monitoring node in real time; The prediction results are stored in the dimensions of management unit-microhabitat-slave monitoring node-time stamp; It should be noted that the role of dividing the management unit and constructing the microhabitat prediction model driven by the main monitoring node is to convert abstract data analysis results (response cluster) into specific management units, achieving precise maintenance at the microhabitat scale, avoiding resource waste, predicting slave monitoring node data through main monitoring node data, without the need to configure high-frequency monitoring equipment for slave monitoring nodes, only periodic calibration is required, further reducing long-term monitoring and operation costs, upgrading passive monitoring to active prediction, providing forward-looking and real-time environmental data support for intelligent garden management, and changing management decisions from experience-driven to data-driven.

[0022] The technical scheme and advantages of the embodiments of the present application are as follows: the environmental parameter values collected by a plurality of sensor nodes in a historical period in the garden and the microtopographic factors are associated and analyzed to determine whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity; if there is, the garden space is divided into a plurality of microhabitats according to the microtopographic factors and environmental characteristics; the monitoring node deployment area is selected by performing representativeness analysis on the microtopographic factors and environmental characteristics in the microhabitats, a master monitoring node and a plurality of slave monitoring nodes are set for each microhabitat, and the garden is monitored; the monitoring data of each master and slave monitoring node are collected and analyzed, the environmental response mode of each master and slave monitoring node is extracted, and each monitoring node is clustered into a plurality of environmental response mode groups according to the environmental response mode; the monitoring nodes belonging to the same environmental response mode group and the corresponding microhabitats are integrated into a management unit, and a microhabitat prediction model driven by the master monitoring node is constructed to support differentiated maintenance decisions. The present application determines whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity by associating and analyzing the environmental parameter values monitored by each sensor node in the garden in a historical period and the microtopographic factors, if there is, the grid unit is divided, and the grid units with similar environmental characteristics and adjacent in space are divided into a microhabitat, and master and slave monitoring nodes are set for each microhabitat, the monitoring nodes are clustered into a plurality of response clusters according to the environmental response mode, the management unit is divided, and a microhabitat prediction model driven by the master monitoring node is constructed, the data of the slave monitoring nodes are predicted according to the data of the master monitoring nodes, the problem of insufficient representativeness of single-point monitoring data caused by microtopographic heterogeneity in the garden is solved, and comprehensive perception and differentiated accurate management of the garden environment with low cost and high precision are realized.

[0023] Embodiment 2: please refer to Figure 3 The garden intelligent management system based on the Internet of Things provided by the embodiments of the present application comprises the following modules: The problem determination module: the environmental parameter values collected by a plurality of sensor nodes in a historical period in the garden and the microtopographic factors are associated and analyzed to determine whether there is a problem of insufficient representativeness of monitoring data caused by microtopographic heterogeneity; The microhabitat division module: if there is, the garden space is divided into a plurality of microhabitats according to the microtopographic factors and environmental characteristics; The monitoring node setting module: the monitoring node deployment area is selected by performing representativeness analysis on the microtopographic factors and environmental characteristics in the microhabitats, a master monitoring node and a plurality of slave monitoring nodes are set for each microhabitat, and the garden is monitored; The response mode group division module: the monitoring data of each master and slave monitoring node are collected and analyzed, the environmental response mode of each master and slave monitoring node is extracted, and each monitoring node is clustered into a plurality of environmental response mode groups according to the environmental response mode; The prediction model construction module integrates the monitoring nodes belonging to the same environment response mode group and the corresponding microhabitats into a management unit, and constructs a microhabitat prediction model driven by a main monitoring node, to support differentiated maintenance decisions.

[0024] The above has carried out the detailed explanation to one embodiment of the application, but the content is only the preferred embodiment of the application, cannot be considered for limiting the implementation scope of the application. Any equivalent changes and improvements made within the scope of the application are still within the scope of the application.

Claims

1. A smart garden management method based on the Internet of Things, characterized in that: Includes the following steps: Correlation analysis was performed on environmental parameter values ​​and micro-topographic factors collected by multiple sensor nodes in the garden during the historical period to determine whether micro-topographic heterogeneity led to insufficient representativeness of the monitoring data. If present, the garden space is divided into multiple microhabitats based on micro-topographic factors and environmental characteristics; By conducting a representative analysis of micro-topographic factors and environmental characteristics within micro-habitats, monitoring node deployment areas were selected. For each micro-habitat, a main monitoring node and multiple secondary monitoring nodes were set up to monitor the garden. The monitoring data of each master and slave monitoring node is collected and analyzed. The environmental response patterns of each master and slave monitoring node are extracted and clustered into multiple environmental response pattern groups according to the environmental response patterns. The monitoring nodes belonging to the same environmental response mode group and their corresponding microhabitats are integrated into a management unit, and a microhabitat prediction model driven by the main monitoring node is constructed to support differentiated maintenance decisions.

2. The intelligent garden management method based on the Internet of Things according to claim 1, characterized in that: The process of the correlation analysis is as follows: Obtain management data for the garden throughout its historical period, including: Micro-topographic factors: elevation, slope, aspect, surface curvature, surface roughness; The environmental parameters monitored by each sensor node include soil moisture, soil temperature, air temperature and humidity, and light intensity. The spatial location information of each node must also be included. Calculate the coefficient of variation of environmental parameters within a 5m radius of the sensor node, and mark the sensor nodes whose coefficient of variation meets the requirements as discrete nodes; A distance weight matrix is ​​constructed based on the coordinates of each sensor node, and the Moran index is calculated according to the Moran index calculation formula. If the Moran index is less than or equal to 0, then the garden has micro-topographical heterogeneity. For any given sensor node: Calculate the Pearson correlation coefficient between the environmental parameter variation coefficient and the micro-topographic factor value. If the Pearson correlation coefficient meets the requirements, it is considered a strong correlation.

3. The intelligent garden management method based on the Internet of Things according to claim 2, characterized in that: The method for determining whether micro-topographic heterogeneity leads to insufficient representativeness of monitoring data is as follows: The proportion of discrete nodes among all sensor nodes is calculated to obtain the proportion of discrete nodes. If the proportion of discrete nodes meets the requirements, the garden exhibits micro-topographical heterogeneity, and the coefficient of variation of environmental parameters is strongly correlated with at least one micro-topographical factor, then there is a problem of insufficient representativeness of monitoring data due to micro-topographical heterogeneity.

4. The intelligent garden management method based on the Internet of Things according to claim 1, characterized in that: The method of dividing the garden space into multiple microhabitats is as follows: The garden is divided into multiple grid units. The micro-topographic factors and environmental parameters of each grid unit are standardized. Spatially adjacent grid units are first grouped into candidate groups. Then, feature similarity between grid units is calculated within the candidate groups to perform feature clustering. The final clustered unit is the micro-habitat.

5. The intelligent garden management method based on the Internet of Things according to claim 1, characterized in that: The method for selecting the deployment area of ​​the monitoring nodes is as follows: Calculate the mean values ​​of micro-topographic factors and environmental parameters for each microhabitat, conduct a field survey of the entire microhabitat area, exclude areas where the micro-topographic factors exceed the microhabitat mean and the environmental parameters exceed the microhabitat environmental parameter mean, and the remaining areas are candidate areas; Three measurement points are evenly distributed in each candidate area to measure the mean value of micro-topographic factors and environmental parameters, and the average value is calculated to obtain the mean value of the candidate area. Calculate the relative deviation between the mean of the candidate area and the mean of the microhabitat. If the relative deviations of all microtopographic factors and environmental parameters meet the requirements, then the area is selected as the main monitoring node deployment area.

6. The intelligent garden management method based on the Internet of Things according to claim 5, characterized in that: The method for selecting the deployment area of ​​the monitoring nodes is also as follows: The mean value of micro-topographic factors of micro-habitat is calculated as the micro-topographic baseline value. The deviation of micro-topographic factors of each grid cell in the micro-habitat from the micro-topographic baseline value is calculated. Grid cells with deviations that meet the requirements are marked as micro-topographic difference areas. The mean values ​​of environmental parameters of microhabitat are calculated as environmental baseline values. Environmental parameters are measured in micro-topographic difference areas. The deviations of environmental parameters of each grid cell in the micro-topographic difference area from the environmental parameter baseline values ​​are calculated. Grid cells with deviations that meet the baseline values ​​are marked as environmental difference areas. In each area of ​​environmental difference, multiple measurement points are set up to measure micro-topographic factors and environmental parameters; Calculate the mean values ​​of the parameters at the measurement points, select the point closest to the mean value of the region as the secondary monitoring node, and ensure that the straight-line distance between the secondary monitoring node and the primary monitoring node does not exceed the maximum radius of the microhabitat.

7. The intelligent garden management method based on the Internet of Things according to claim 1, characterized in that: The method for extracting the environmental response modes of each master and slave monitoring node is as follows: From the time-series monitoring data of each master and slave monitoring node, daily variation patterns, seasonal variation trends, extreme weather response characteristics, and external correlation characteristics are extracted, specifically including: Diurnal variation characteristics: the diurnal fluctuation range of parameters, the time of peak occurrence, and the time of trough occurrence; Seasonal variation trend characteristics: seasonal rate of increase, rate of decrease, and seasonal mean of parameters; Extreme weather response characteristics: Response amplitude = |parameter extreme value after extreme weather - parameter baseline value before extreme weather|, response rate = response amplitude / time interval from the start of extreme weather to the parameter reaching the extreme value; External correlation characteristics: Pearson correlation coefficients between nodal parameters and external climate factors.

8. The intelligent garden management method based on the Internet of Things according to claim 7, characterized in that: The method for clustering each monitoring node into multiple environmental response mode groups is as follows: The four types of environmental response features—diurnal variation patterns, seasonal variation trends, extreme weather response features, and external correlation features—are standardized and used as clustering features to obtain multiple environmental response model groups.

9. The intelligent garden management method based on the Internet of Things according to claim 1, characterized in that: The construction and prediction method of the microhabitat prediction model driven by the main monitoring node is as follows: Integrate the master and slave monitoring nodes and their respective microhabitats within the same environmental response mode group into a single management unit; The monitoring data from the master monitoring node and the slave monitoring node are integrated into a feature set, specifically including: Core driving features: real-time values, rate of change, and cumulative values ​​of key parameters of the main monitoring node; Spatial correlation characteristics: based on the straight-line distance between the monitoring node and the main monitoring node, the elevation difference, the aspect difference, and whether they belong to the same response cluster; Environmental background characteristics: soil type, ecological unit type, average annual sunshine duration, soil organic matter content, and land cover type; The monitoring data is aligned by timestamp, and a corresponding feature set is matched for each monitoring node to form a complete training dataset. Random forest was used to train the model with the training dataset to obtain a microhabitat prediction model driven by the main monitoring node. The main monitoring node collects data in real time, generates core driving features, retrieves spatial correlation features and environmental background features from the secondary monitoring nodes, and inputs them into the microhabitat prediction model driven by the main monitoring node. The model outputs the predicted values ​​of key parameters from the secondary monitoring nodes in real time.

10. A smart garden management system based on the Internet of Things, characterized in that: Includes the following modules: Problem identification module: Performs correlation analysis on environmental parameter values ​​and micro-topographic factors collected by multiple sensor nodes in the garden during the historical period to determine whether there is a problem of insufficient representativeness of monitoring data due to micro-topographic heterogeneity; Microhabitat segmentation module: If it exists, the garden space is divided into multiple microhabitats based on micro-topographic factors and environmental characteristics; Monitoring node setting module: By conducting representative analysis of micro-topographic factors and environmental characteristics within micro-habitats, the deployment area of ​​monitoring nodes is selected. One main monitoring node and multiple slave monitoring nodes are set for each micro-habitat to monitor the garden. Response mode grouping module: Collects and analyzes monitoring data from each master and slave monitoring node, extracts the environmental response mode of each master and slave monitoring node, and clusters each monitoring node into multiple environmental response mode groups based on the environmental response mode; Predictive model building module: Integrates monitoring nodes belonging to the same environmental response mode group and their corresponding microhabitats into a management unit, and builds a microhabitat prediction model driven by the main monitoring node to support differentiated maintenance decisions.

Images