Smart garden management method and system based on big data

Abstract

The invention discloses an intelligent garden management method and system based on big data, and belongs to the technical field of garden management. Microclimate region division is carried out on a garden through a clustering algorithm, and the region representative weight of each sensor is calculated in combination with a data heterogeneity response factor and a microtopography interference factor; weighted fusion of sensor data is realized, and a more representative environment estimation value is generated; according to the method, prediction and anomaly detection are carried out on a plant growth environment trend based on time sequence analysis, a garden regulation and control instruction is automatically generated, actual plant state feedback is introduced to realize dynamic optimization and adjustment of a weight, and the method effectively improves the accuracy of environment perception and the intelligent level of decision response, and improves the user experience. And resource misuse and plant health risks caused by information distortion are obviously reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of garden management, and in particular to a smart garden management method and system based on big data. Background Art

[0002] Smart garden management refers to the intelligent, digital management of the entire process of landscaping planning, construction, maintenance, and management, leveraging advanced information technologies such as the Internet of Things, big data, and artificial intelligence. By using sensors to collect real-time environmental data (such as temperature and humidity, soil moisture, and pests and diseases), combined with intelligent analysis and remote control, this allows for precise regulation of plant growth environments, rational allocation of maintenance resources, and scientific management of garden landscapes. This improves garden management efficiency and ecological benefits, driving the development of landscaping towards intelligent and sustainable development.

[0003] The existing technology has the following shortcomings:

[0004] In smart garden management, microclimate differences can distort model input data, leading to misjudgment. If sensors are unevenly distributed, they cannot fully reflect the true environmental conditions within the park. The model may generalize local data into overall trends, misleading management decisions. For example, in an eco-park, the south side is low-lying and prone to waterlogging, but sensors are concentrated on the drier north side. This can cause the system to misjudge the overall "soil dryness" in the park and automatically initiate irrigation, exacerbating waterlogging on the south side and causing plant root diseases. Summary of the Invention

[0005] The purpose of the present invention is to provide a smart garden management method and system based on big data to address the shortcomings of the background technology.

[0006] In order to achieve the above-mentioned purpose, the present invention provides the following technical solution: a smart garden management method based on big data, comprising:

[0007] The garden area is divided into several microclimate sub-areas. By acquiring and analyzing the historical environmental data and topographic information in different sub-areas, a clustering algorithm is used to construct a microclimate model of the park.

[0008] Based on the microclimate model, a regional weight parameter is assigned to each sensor node, wherein the weight parameter is calculated based on the data heterogeneity response factor and the micro-topography interference factor of the node;

[0009] Performing weighted fusion processing on the raw environmental data collected by the sensors to obtain a weighted environmental estimation value, wherein the weighted processing includes multiplying each sensor data with its corresponding weight parameter and summing the result;

[0010] The weighted environmental estimates obtained within a fixed time period are used to construct a time series for input into a trend prediction model to evaluate plant growth environment trends and detect anomalies.

[0011] Based on the evaluation results, garden management control instructions are automatically generated, and combined with the actual plant status feedback data for analysis, the regional weight parameters are optimized according to the analysis results.

[0012] Preferably, constructing a microclimate model by acquiring and analyzing historical environmental data and terrain information in different sub-regions includes:

[0013] The system collects topographic data, soil property information, and historical sensor environmental monitoring data from the garden area, including temperature, humidity, wind speed, and light intensity. The garden area is divided into several basic unit areas. For each basic unit area, a feature vector is constructed, including the average temperature variation, humidity fluctuation, slope value, and historical NDVI mean. All feature vectors are normalized and input into a clustering algorithm for clustering, forming a garden area microclimate sub-area layer.

[0014] Preferably, the calculation of the regional weight parameter includes:

[0015] The data heterogeneity response factor and micro-topography disturbance factor were constructed as a combined feature vector and input into the polynomial regression model;

[0016] The model is trained with the goal of minimizing the prediction error of all sensor node weights;

[0017] The regional weight parameters of each sensor node are determined using the output results of the trained model.

[0018] Preferably, the method for obtaining the data heterogeneity response factor SVC is as follows: the current observation data of the input node i is a multidimensional vector x i , the adjacent node set of node i is Contains N neighbor nodes; the observation data of each neighbor node j is a vector x of the same dimension j ;Construct adjacent node data matrix Each row is the observation vector x of the adjacent node j , dimension d is the number of features, R is a set of real numbers; calculate the mean vector μ∈R of the neighborhood data d , the expression is: Calculate the neighborhood covariance matrix Σ∈R d×d , the expression is: Where T is the matrix transpose, and the data heterogeneity response factor SVC of the target node i is calculated as follows: Σ -1represents the inverse of the covariance matrix.

[0019] Preferably, the method for obtaining the micro-topography interference factor TII is: obtaining high-resolution DEM data of the park, deriving the slope layer β from the DEM, and representing the slope of each pixel; the slope calculation expression is: in is the gradient of elevation in the x and y directions; use the D8 flow direction algorithm or the multi-flow direction algorithm to calculate the upslope catchment area a of each pixel; use the generated slope and catchment area grids to substitute the formula point by point to calculate the terrain wetness index TWI, which is expressed as: The normalized TWI value is used as the micro-topography interference factor TII, and the expression is: Where min(TWI) and max(TWI) represent the minimum and maximum values ​​of the terrain wetness index TWI, respectively.

[0020] Preferably, the weighted fusion process includes:

[0021] At each time point t, the original environmental observations of all sensors are collected; the observation value of each node is multiplied by its corresponding regional weight parameter; all product results are summed to obtain the weighted environmental estimate value at time point t.

[0022] Preferably, the trend prediction model is an LSTM model, the construction of which includes:

[0023] Takes as input a sequence of weighted environmental estimates for consecutive time periods;

[0024] Perform supervised training on the model to enable it to predict environmental estimates for several time steps into the future;

[0025] Determine whether there are abnormal changes in the plant's environment based on the changing trend of the predicted value;

[0026] If the predicted value continuously decreases or increases and exceeds the preset threshold, the anomaly detection flag is triggered.

[0027] Preferably, the step of optimizing the regional weight parameters according to the actual plant status feedback data includes:

[0028] Obtain the current feedback status indicators of plants, including NDVI, chlorophyll content or water content;

[0029] Compare the actual state of the plant with the ideal reference value and calculate the deviation of the plant state;

[0030] Assign error attribution factors based on the degree of deviation between sensor node data and weighted estimates;

[0031] The regional weight parameters are iteratively adjusted using a negative feedback correction formula.

[0032] The present invention also provides a smart garden management system based on big data, which includes a microclimate modeling module, a regional weight calculation module, an environmental data fusion module, an anomaly detection module and an instruction generation module;

[0033] Microclimate modeling module: This module divides the garden area into several microclimate sub-regions. By acquiring and analyzing historical environmental data and terrain information in different sub-regions, a clustering algorithm is used to construct a microclimate model of the park.

[0034] Regional weight calculation module: Based on the microclimate model, a regional weight parameter is assigned to each sensor node. The weight parameter is calculated based on the node's data heterogeneity response factor and micro-topography interference factor;

[0035] Environmental data fusion module: performs weighted fusion processing on the raw environmental data collected by the sensors to obtain a weighted environmental estimate. The weighted processing includes multiplying each sensor data with its corresponding weight parameter and summing the results.

[0036] Anomaly detection module: constructs a time series of weighted environmental estimates obtained within a fixed time period, which is used as input into the trend prediction model for plant growth environment trend assessment and anomaly detection;

[0037] Instruction generation module: Automatically generates garden management and control instructions based on the evaluation results, analyzes the feedback data of the actual plant status, and optimizes the regional weight parameters based on the analysis results.

[0038] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0039] 1. This invention achieves high-precision weighted fusion and dynamic optimization of environmental perception data through microclimate regional modeling, data heterogeneity analysis, and terrain interference perception, significantly improving the system's adaptability to complex landscape ecosystems. Compared to existing technologies, this method effectively addresses environmental misjudgments caused by uneven sensor distribution and microclimate variations, avoiding misleading decisions based on partial generalizations. It ensures more precise and reasonable control actions such as irrigation and lighting, reducing the risk of disease and resource waste.

[0040] 2. This invention further incorporates feedback from actual plant growth status, constructing a closed-loop self-learning mechanism to continuously optimize perception weight parameters. This enables the system to have data-driven adaptive adjustment capabilities, enhancing its predictive sensitivity and responsiveness to environmental changes. The overall solution integrates multi-source heterogeneous data analysis, machine learning modeling, and intelligent control decision-making, not only enhancing the scientific and intelligent level of smart garden management, but also providing an efficient and scalable solution for a variety of application scenarios, including urban green spaces, ecological parks, and agricultural and forestry plantings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0042] Figure 1 This is a mind map of the method of the present invention.

[0043] Figure 2 This is a mind map of the system modules of the present invention. DETAILED DESCRIPTION

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0045] Example 1, please refer to Figure 1 As shown, the smart garden management method based on big data described in this embodiment includes:

[0046] The garden area is divided into several microclimate sub-areas. By acquiring and analyzing the historical environmental data and topographic information in different sub-areas, a clustering algorithm is used to construct a microclimate model of the park.

[0047] Based on the microclimate model, a regional weight parameter is assigned to each sensor node, wherein the weight parameter is calculated based on the data heterogeneity response factor and the micro-topography interference factor of the node;

[0048] Performing weighted fusion processing on the raw environmental data collected by the sensors to obtain a weighted environmental estimation value, wherein the weighted processing includes multiplying each sensor data with its corresponding weight parameter and summing the result;

[0049] The weighted environmental estimates obtained within a fixed time period are used to construct a time series for input into a trend prediction model to evaluate plant growth environment trends and detect anomalies.

[0050] Based on the evaluation results, garden management control instructions are automatically generated, and combined with the actual plant status feedback data for analysis, the regional weight parameters are optimized according to the analysis results.

[0051] To address the problems of significant microclimate differences, uneven sensor distribution, and inaccurate model predictions in garden management, this embodiment divides the park into microclimate sub-regions and establishes a regional representative structure model to support subsequent data weighted fusion, trend prediction, and management decision optimization.

[0052] Collect basic information about the garden area. Data types include, but are not limited to: topographic data (e.g., elevation, slope, and aspect); historical environmental data (e.g., temperature, humidity, wind speed, rainfall, and light intensity); soil properties (e.g., texture type, permeability, and water retention); GPS coordinate distribution and existing sensor locations. Historical data should span at least one complete climate cycle (e.g., 12 months). Sampling intervals can range from 15 minutes to one hour, depending on the device configuration.

[0053] For each basic unit area (e.g., a 10m x 10m grid), a feature vector is constructed, including: mean temperature variation; humidity extremes and standard deviations; topographic factors (e.g., slope); historical mean NDVI index; and soil water-holding capacity. Z-score or Min-Max normalization algorithms are used to unify dimensions and eliminate unit interference.

[0054] Based on the above eigenvectors, a spatial clustering algorithm is used to divide the microclimate sub-regions. Optional clustering algorithms include K-Means, DBSCAN, and GMM (Gaussian Mixture Model). The number of clusters is preset based on the park area and the degree of microclimate difference, such as K = 6. Considering the spatial continuity of the neighborhood, spatial distance constraints and attribute similarity are integrated to improve clustering effectiveness. The clustering results are output in the form of labels to form a microclimate sub-region layer. The clustering results are spatially mapped with the actual deployment points of the sensors to construct a three-layer mapping model of "node-region-representativeness". The specific operations are as follows:

[0055] Each sensor belongs to its cluster area;

[0056] Statistics on the area, vegetation type, water distribution, etc. of each region are used for subsequent representative weight calculation;

[0057] The microclimate model can be dynamically updated to adapt to seasonal changes or sensor deployment adjustments.

[0058] By introducing a microclimate sub-region division mechanism, the present invention enables the system to more accurately identify the environmental characteristics of different regions, improves the spatial rationality of data modeling, avoids the misjudgment problem of sensor points due to "generalizing from a single case", and provides a more reliable modeling basis for subsequent trend prediction and anomaly detection.

[0059] In smart garden management systems, sensor data is the core foundation for environmental modeling and decision-making. However, differences in spatial location, microclimate conditions, and terrain characteristics among different nodes lead to inconsistent regional representation of their observations. Indiscriminately averaging all sensor data can easily lead to trend prediction errors and misleading management.

[0060] To improve modeling accuracy, the system assigns regional representative weight parameters to each sensor node based on the established microclimate model, which serves as an important basis for weighted environmental estimation and input model.

[0061] Through historical meteorological data and terrain analysis, the park has been divided into several microclimate sub-regions (such as moist woodlands, sunny slope grasslands, low-lying waterfront areas, etc.), and the environmental conditions in each sub-region are relatively homogeneous.

[0062] The weight calculation of each node depends on the following data:

[0063] Historical environmental monitoring data (temperature, humidity, light, wind speed, etc.);

[0064] Environmental data of neighboring nodes (for data heterogeneity analysis);

[0065] Digital elevation model (DEM) and terrain-derived factors (slope, aspect, curvature, etc.);

[0066] The relative position of the node within the region (center / edge).

[0067] The data heterogeneity response factor SVC is used to indicate the degree of fluctuation difference between the monitoring data of a certain node and its neighboring nodes within a certain time window, reflecting its data representativeness or local deviation risk.

[0068] The acquisition method is: input node i’s current observation data as a multidimensional vector x i , including soil moisture, air temperature, light intensity, wind speed, NDVI, etc.; the adjacent node set of node i is Contains N neighbor nodes; the observation data of each neighbor node j is a vector x of the same dimension j .

[0069] Construct adjacent node data matrix Each row is the observation vector x of the adjacent node j , dimension d is the number of features, and R is a set of real numbers;

[0070] Calculate the mean vector μ∈R of the neighborhood data d , the expression is: Calculate the neighborhood covariance matrix Σ∈R d×d , the expression is: Where T is the matrix transpose, and the data heterogeneity response factor SVC of the target node i is calculated as follows: Σ -1 Represents the inverse matrix of the covariance matrix. If SVC>3 (or based on empirical threshold), the node behavior is considered to be highly heterogeneous.

[0071] The micro-topography interference factor TII is used to indicate whether the node location is susceptible to special micro-topography (such as depressions, steep slopes, and extreme directions).

[0072] The acquisition method is as follows: obtain high-resolution DEM (digital elevation model) data of the park, with a recommended grid accuracy of 1–5 meters; use GIS software (such as ArcGIS / QGIS) or Python tools (such as GDAL, WhiteboxTools) to process terrain data.

[0073] The slope layer β is derived from the DEM, which represents the slope of each pixel. The slope can be calculated as: in is the gradient of elevation in the x and y directions;

[0074] Calculate the upslope catchment area a for each pixel using the D8 flow direction algorithm or the multi-flow direction (MFD) algorithm. The corresponding tool in ArcGIS / QGIS is "Flow Accumulation." The output unit is usually the number of pixels, which can be converted to m by multiplying by the pixel area. 2 .

[0075] Using the slope and catchment area grids generated in the first two steps, we substitute them point by point into the formula to calculate the terrain wetness index TWI. The expression is: For regions where tanβ≈0 (eg very flat), a lower limit of tanβmin=0.001 is added to prevent division by zero.

[0076] The normalized TWI value is used as the micro-topography interference factor TII, and the expression is: Where min(TWI) and max(TWI) represent the minimum and maximum values ​​of the terrain wetness index TWI, respectively. TII∈[0,1], where a larger value indicates that the location is more susceptible to wet disturbance and the micro-topography interference is stronger.

[0077] Convert the data heterogeneity response factor and the microtopography interference factor into a comprehensive feature vector, use the comprehensive feature vector as the input of the machine learning model, and use the machine learning model to predict the weight parameter label of each sensor node allocation area with each group of comprehensive feature vectors as the prediction target, and minimize the sum of the prediction errors of the weight parameter labels of each sensor node allocation area as the training target. Train the machine learning model until the sum of the prediction errors converges and then stop the model training. Determine the weight parameters of each sensor node allocation area according to the model output results. Among them, the machine learning model is a polynomial regression model.

[0078] Perform a weighted summation calculation on the obtained weight parameters of each sensor node allocation area and the original environmental data collected by the sensor to obtain a weighted environmental estimation value.

[0079] Set the time resolution (such as every 30 minutes, every hour, every day); calculate the weighted estimation value for a certain environmental variable (such as soil humidity, light, temperature) at each time step t using the aforementioned method.

[0080] Within a fixed time interval, form a time series E. The model selection (depending on the target and accuracy requirements) includes: simple models: moving average, exponential smoothing; statistical models: ARIMA, SARIMA; machine learning models: XGBoost, Random Forest (lag features need to be introduced); deep learning models: LSTM (long short-term memory neural network); GRU (gated recurrent unit) and Temporal Convolutional Network (TCN).

[0081] Input the historical time series E, output the future prediction value or the current trend label (rising / falling / stable); calculate the slope, fluctuation amplitude, and extreme points of the prediction result to evaluate the changing trend of the environment where the plant is located; the trend threshold can be set: rapid decline → drought warning; sharp increase → heat damage or waterlogging risk; Method 1: When the error between the prediction value and the actual value exceeds the set threshold, an anomaly is triggered; Method 2: Combine the confidence interval of the model to judge rare fluctuations.

[0082] If a trend anomaly or a major deviation is detected, the system automatically generates control strategy suggestions, such as: starting or delaying irrigation; issuing an artificial inspection alarm; adjusting the operation of shading or ventilation equipment.

[0083] Compare the predicted value E^t+h with the plant environment threshold interval [Emin, Emax] to determine whether it exceeds the boundary: if E^t+h < Emin, trigger a lower limit warning; if E^t+h > Emax, trigger an upper limit warning. At the same time, the "sharp change" situation can be identified by combining the decline / rise speed threshold.

[0084] Match the corresponding instruction template according to the anomaly type:

[0085] If humidity drops to a level that poses a risk of drought, irrigation is initiated, for example, "Region A starts drip irrigation mode for 30 minutes";

[0086] If the humidity rises to oversaturation, stop irrigation and give a warning of drainage, for example, "stop irrigation in area B; check drainage facilities";

[0087] If the sunlight is insufficient (continuous cloudy days are predicted), the supplementary lighting device is activated, for example, "activate the LED supplementary lighting in area C for 3 hours";

[0088] If the wind speed suddenly increases (typhoon forecast), the sunshade net will be retracted and the awning will be reinforced, for example, "Area D rolls up the sunshade and issues a patrol order."

[0089] Fill in the policy template with the instruction format that can be recognized by the specific system, including the following fields:

[0090] Zone number, control object type (irrigation, ventilation, lighting, patrol, etc.), execution action (on / off / delay, etc.), device ID (optional), execution time or duration, and priority label (high / medium / low).

[0091] The instructions are pushed to the garden control subsystem (via MQTT, Modbus, REST API, etc.) for automatic execution or execution after manual confirmation (depending on the system level settings).

[0092] In smart garden management, regional weight parameters determine the contribution of sensor nodes to overall environmental estimation. Due to factors such as microclimate, topography, and data drift, static weights may not maintain high accuracy over time. Therefore, this invention adaptively optimizes weight parameters based on actual plant status feedback, allowing the system to adjust the perception layer weight configuration based on actual plant responses.

[0093] Plant state feedback error ΔS t Calculation, the expression is: Indicates the reference value that plants should present under ideal conditions, S t Represents the actual plant status feedback value (such as chlorophyll index, NDVI, and water content); larger deviations indicate errors in environmental estimation, which may be caused by misweighting in the perception layer.

[0094] For each node i, the attribution weight (normalized) is calculated based on the possible responsibility of its current value for the overall estimation error, and the expression is: Where, Represents the collected value of node i at the current moment; attribution weight The larger the value is, the greater the contribution of the node to the error is likely to be. tIt represents the environmental estimation value obtained by weighted fusion of all current sensor data, and d is the total number of nodes. It adopts negative feedback adjustment method, and the adjustment direction is opposite to the deviation direction of plant state. The expression is: Where α is the learning rate (control weight adjustment range); represents the weight parameter of sensor node i in the region at time t, represents the weight parameter of sensor node i in the region at time t+1.

[0095] If the plant state is lower than the ideal value (such as yellow leaves or insufficient water), the system considers that the environmental estimate is too high; the node value i that is much higher than the estimated value should be weighted down to reduce its misleading effect; if the node value is low but the environmental preference is misjudged, its weight should be increased (reverse offset).

[0096] Example 2, please refer to Figure 2 As shown, the smart garden management system based on big data described in this embodiment includes a microclimate modeling module, a regional weight calculation module, an environmental data fusion module, an anomaly detection module and an instruction generation module;

[0097] Microclimate modeling module: This module divides the garden area into several microclimate sub-regions. By acquiring and analyzing historical environmental data and terrain information in different sub-regions, a clustering algorithm is used to construct a microclimate model of the park.

[0098] Regional weight calculation module: Based on the microclimate model, a regional weight parameter is assigned to each sensor node. The weight parameter is calculated based on the node's data heterogeneity response factor and micro-topography interference factor;

[0099] Environmental data fusion module: performs weighted fusion processing on the raw environmental data collected by the sensors to obtain a weighted environmental estimate. The weighted processing includes multiplying each sensor data with its corresponding weight parameter and summing the results.

[0100] Anomaly detection module: constructs a time series of weighted environmental estimates obtained within a fixed time period, which is used as input into the trend prediction model for plant growth environment trend assessment and anomaly detection;

[0101] Instruction generation module: Automatically generates garden management and control instructions based on the evaluation results, analyzes the feedback data of the actual plant status, and optimizes the regional weight parameters based on the analysis results.

[0102] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0103] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0104] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0105] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A smart garden management method based on big data, characterized by: include: The garden area is divided into several microclimate sub-areas. By acquiring and analyzing the historical environmental data and topographic information in different sub-areas, a clustering algorithm is used to construct a microclimate model of the park. Based on the microclimate model, a regional weight parameter is assigned to each sensor node, wherein the weight parameter is calculated based on the data heterogeneity response factor and the micro-topography interference factor of the node; Performing weighted fusion processing on the raw environmental data collected by the sensors to obtain a weighted environmental estimation value, wherein the weighted processing includes multiplying each sensor data with its corresponding weight parameter and summing the result; The weighted environmental estimates obtained within a fixed time period are used to construct a time series for input into a trend prediction model to evaluate plant growth environment trends and detect anomalies. Based on the evaluation results, garden management control instructions are automatically generated, and combined with the actual plant status feedback data for analysis, the regional weight parameters are optimized according to the analysis results.

2. The method for intelligent garden management based on big data according to claim 1, characterized in that: The construction of the microclimate model by acquiring and analyzing historical environmental data and terrain information in different sub-regions includes: The system collects topographic data, soil property information, and historical sensor environmental monitoring data from the garden area, including temperature, humidity, wind speed, and light intensity. The garden area is divided into several basic unit areas. For each basic unit area, a feature vector is constructed, including the average temperature variation, humidity fluctuation, slope value, and historical NDVI mean. All feature vectors are normalized and input into a clustering algorithm for clustering, forming a garden area microclimate sub-area layer.

3. The big data-based smart garden management method according to claim 1, characterized in that: The calculation of the regional weight parameter includes: The data heterogeneity response factor and micro-topography disturbance factor were constructed as a combined feature vector and input into the polynomial regression model; The model is trained with the goal of minimizing the prediction error of all sensor node weights; The regional weight parameters of each sensor node are determined using the output results of the trained model.

4. The big data-based smart garden management method according to claim 3, characterized in that: The method for obtaining the data heterogeneity response factor SVC is as follows: the current observation data of input node i is a multidimensional vector x i , the adjacent node set of node i is Contains N neighbor nodes; the observation data of each neighbor node j is a vector x of the same dimension j ;Construct adjacent node data matrix Each row is the observation vector x of the adjacent node j , dimension d is the number of features, R is a set of real numbers; calculate the mean vector μ∈R of the neighborhood data d , the expression is: Calculate the neighborhood covariance matrix Σ∈R d×d , the expression is: μ) T ; Where T is the matrix transpose, and the data heterogeneity response factor SVC of the target node i is calculated as follows: Σ -1 represents the inverse of the covariance matrix.

5. The big data-based smart garden management method according to claim 4, characterized in that: The method for obtaining the micro-topography interference factor TII is to obtain high-resolution DEM data of the park, derive the slope layer β from the DEM, and represent the slope of each pixel; the slope calculation expression is: in is the gradient of elevation in the x and y directions; use the D8 flow direction algorithm or the multi-flow direction algorithm to calculate the upslope catchment area a of each pixel; use the generated slope and catchment area grids to substitute the formula point by point to calculate the terrain wetness index TWI, which is expressed as: The normalized TWI value is used as the micro-topography interference factor TII, and the expression is: Where min(TWI) and max(TWI) represent the minimum and maximum values ​​of the terrain wetness index TWI, respectively.

6. The big data-based smart garden management method according to claim 5, characterized in that: The weighted fusion process includes: At each time point t, the original environmental observations of all sensors are collected; the observation value of each node is multiplied by its corresponding regional weight parameter; all product results are summed to obtain the weighted environmental estimate value at time point t.

7. The big data-based smart garden management method according to claim 6, characterized in that: The trend prediction model is an LSTM model, and its construction includes: Takes as input a sequence of weighted environmental estimates for consecutive time periods; Perform supervised training on the model to enable it to predict environmental estimates for several time steps into the future; Determine whether there are abnormal changes in the plant's environment based on the changing trend of the predicted value; If the predicted value continuously decreases or increases and exceeds the preset threshold, the anomaly detection flag is triggered.

8. The big data-based smart garden management method according to claim 7, characterized in that: The step of optimizing the regional weight parameters according to the actual plant status feedback data includes: Obtain the current feedback status indicators of plants, including NDVI, chlorophyll content or water content; Compare the actual state of the plant with the ideal reference value and calculate the deviation of the plant state; Assign error attribution factors based on the degree of deviation between sensor node data and weighted estimates; The regional weight parameters are iteratively adjusted using a negative feedback correction formula.

9. A smart garden management system based on big data, used to implement the smart garden management method based on big data according to any one of claims 1 to 8, characterized in that: It includes microclimate modeling module, regional weight calculation module, environmental data fusion module, anomaly detection module and instruction generation module; Microclimate modeling module: This module divides the garden area into several microclimate sub-regions. By acquiring and analyzing historical environmental data and terrain information in different sub-regions, a clustering algorithm is used to construct a microclimate model of the park. Regional weight calculation module: Based on the microclimate model, a regional weight parameter is assigned to each sensor node. The weight parameter is calculated based on the node's data heterogeneity response factor and micro-topography interference factor; Environmental data fusion module: performs weighted fusion processing on the raw environmental data collected by the sensors to obtain a weighted environmental estimate. The weighted processing includes multiplying each sensor data with its corresponding weight parameter and summing the results. Anomaly detection module: constructs a time series of weighted environmental estimates obtained within a fixed time period, which is used as input into the trend prediction model for plant growth environment trend assessment and anomaly detection; Instruction generation module: Automatically generates garden management and control instructions based on the evaluation results, analyzes the feedback data of the actual plant status, and optimizes the regional weight parameters based on the analysis results.

Images