Intelligent monitoring and management system for urban garden irrigation

Abstract

The invention discloses an intelligent monitoring and management system for urban garden irrigation. Initial multi-dimensional garden data are collected through a sensor and meteorological equipment; extracting associated features in the initial multi-dimensional garden data by using a machine learning algorithm, establishing a Bi-LSTM garden dynamic water demand prediction model based on a Bi-LSTM neural network, and optimizing hyper-parameters of the prediction model by using a PSO particle swarm optimization algorithm to obtain a target Bi-LSTM garden dynamic water demand prediction model; and inputting the multi-dimensional associated garden data into the target B i-LSTM garden dynamic water demand prediction model for training, optimizing a water-fertilizer ratio through an SVM support vector machine algorithm to obtain a garden zoning irrigation strategy, and irrigating the garden based on the garden zoning irrigation strategy. The future water demand of the garden can be accurately predicted, the growth of plants is prevented from being affected by water shortage or excessive water, and healthy growth of the garden plants is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of garden irrigation, in particular to an intelligent monitoring and management system for urban garden irrigation. Background Art

[0002] With the acceleration of urbanization and the continuous expansion of urban garden areas, garden irrigation, as a key link in maintaining the normal growth of garden plants, has become increasingly important. Traditional garden irrigation methods mainly include manual irrigation and timed irrigation. Manual irrigation relies on manual operation and suffers from problems such as low irrigation efficiency, high labor intensity, and severe water waste. While timed irrigation can achieve a certain degree of automation, it does not consider actual soil moisture, meteorological conditions, and plant water requirements, which can easily lead to under-irrigation or over-irrigation, affecting plant growth and causing inappropriate use of water resources.

[0003] In recent years, a number of intelligent irrigation systems have emerged. Most of these systems use sensors to collect soil moisture data and then control irrigation equipment based on preset thresholds. However, these systems have the following shortcomings: First, they rely solely on single soil moisture data and lack comprehensive consideration of multiple factors, such as meteorological conditions and plant growth stages, making it difficult to achieve precise irrigation. Second, the monitoring and management functions are relatively independent, preventing effective closed-loop management and delaying the timely handling of abnormal situations during the irrigation process. Third, the system's limited intelligence makes it impossible to dynamically adjust irrigation strategies based on actual conditions, failing to meet the diverse and personalized irrigation needs of urban gardens. Therefore, there is an urgent need for an urban garden irrigation system that can comprehensively consider multiple factors and achieve integrated intelligent monitoring and management. Summary of the Invention

[0004] The purpose of the present invention is to solve the above problems and to design an intelligent monitoring and management system for urban garden irrigation.

[0005] The technical solution of the present invention to achieve the above object is that, further, in the above-mentioned urban garden irrigation intelligent monitoring and management system, the urban garden irrigation intelligent monitoring and management system includes:

[0006] The garden data acquisition module is used to collect soil temperature and humidity data, garden meteorological data, and plant growth detection data in the garden through sensors and meteorological equipment to obtain initial multi-dimensional garden data;

[0007] A data processing and analysis module is used to extract correlation features from the initial multi-dimensional garden data using a machine learning algorithm to obtain multi-dimensional correlated garden data;

[0008] The prediction model building module is used to establish a Bi-LSTM garden dynamic water demand prediction model based on the Bi-LSTM neural network, and use the PSO particle swarm optimization algorithm to optimize the hyperparameters of the prediction model to obtain the target Bi-LSTM garden dynamic water demand prediction model;

[0009] A water demand prediction module is used to input the multi-dimensional associated garden data into the target Bi-LSTM garden dynamic water demand prediction model for training to obtain the garden dynamic water demand;

[0010] The garden zoning irrigation module is used to divide the garden into multiple irrigation units based on the dynamic water demand of the garden, set differentiated irrigation thresholds according to the irrigation units, optimize the water-fertilizer ratio through the SVM support vector machine algorithm, obtain the garden zoning irrigation strategy, and irrigate the garden based on the garden zoning irrigation strategy.

[0011] Furthermore, in the above-mentioned urban garden irrigation intelligent monitoring and management system, the garden data acquisition module includes the following submodules:

[0012] The soil temperature and humidity data acquisition submodule is used to collect temperature and humidity data in the soil using temperature and humidity sensors to obtain soil temperature and humidity data;

[0013] The garden meteorological data collection submodule is used to monitor the garden meteorological data in real time using an anemometer, a rain gauge, a temperature and humidity sensor, and a light intensity sensor, including at least wind speed, rainfall, temperature, humidity, image data, and light intensity;

[0014] The plant growth detection data collection submodule is used to set corresponding growth monitoring sensors for different types of garden plants, including at least leaf moisture sensors and stem flow sensors, to collect plant growth detection data;

[0015] The data matching submodule is used to synchronize timestamps and match spatial coordinates of multi-source data collected by sensors using the Kalman filter algorithm, eliminating sensor drift errors and obtaining multi-dimensional garden data.

[0016] The data processing submodule is used to identify outliers in the multi-dimensional garden data based on the Isolation Forest algorithm, and fill in missing data using the neighboring sensor interpolation method to obtain initial multi-dimensional garden data.

[0017] Furthermore, in the above-mentioned urban garden irrigation intelligent monitoring and management system, the data processing and analysis module includes the following submodules:

[0018] A time series feature extraction submodule is used to calculate the soil moisture change rate, the cumulative difference between evaporation and rainfall in the initial multi-dimensional garden data through a sliding window, obtain the impact of short-term environmental fluctuations on water demand, and obtain time series feature data;

[0019] The spatial feature extraction submodule is used to perform semantic segmentation on image data using the CNN convolutional neural network, extract the spatial distribution characteristics of vegetation density, canopy coverage area and terrain slope, and obtain spatial feature data;

[0020] The cross-feature extraction submodule is used to calculate the influence weight of the interaction between soil salinity and pH value on plant water absorption efficiency based on random forest, establish soil and plant coupling characteristics, and obtain cross-feature data;

[0021] The feature data enhancement submodule is used to simulate the soil moisture change curve under different meteorological conditions based on the GAN generative adversarial network, expand the small sample scene data, fuse the extracted feature data, and obtain multi-dimensional correlated garden data.

[0022] Furthermore, in the above-mentioned urban garden irrigation intelligent monitoring and management system, the prediction model establishment module includes: establishing a Bi-LSTM garden dynamic water demand prediction model based on the Bi-LSTM neural network. The Bi-LSTM network synchronously processes time series data through forward and backward LSTM layers to capture irrigation demand and historical meteorological events; the backward layer of the prediction model traces back the soil moisture attenuation trend over the past three days, and the forward layer predicts the transpiration rate based on the light intensity in the next six hours.

[0023] Furthermore, in the above-mentioned urban garden irrigation intelligent monitoring and management system, the prediction model establishment module also includes the following submodules:

[0024] The optimization submodule is used to optimize the hyperparameters of the Bi-LSTM garden dynamic water demand prediction model using the particle swarm optimization (PSO) algorithm;

[0025] The optimization submodule is used to simulate the foraging behavior of bird flocks based on the PSO particle swarm algorithm. It searches for the optimal hyperparameter combination through continuous iterative search of particles in the solution space, including at least the number of network layers, number of neurons, and learning rate.

[0026] The submodule is obtained, which is used to adjust the hyperparameters with the goal of minimizing the prediction error during the optimization process, and obtain the target Bi-LSTM garden dynamic water demand prediction model.

[0027] Furthermore, in the above-mentioned urban garden irrigation intelligent monitoring and management system, the garden zoning irrigation module includes: dividing the garden into multiple irrigation units according to the garden's topography, plant species distribution, soil type factors, and combined with the garden's dynamic water demand, and dividing the water-loving plant area and the drought-resistant plant area into different irrigation units respectively.

[0028] Furthermore, in the above-mentioned urban garden irrigation intelligent monitoring and management system, the garden zone irrigation module includes the following submodules:

[0029] The threshold submodule is used to obtain the plant growth stage, soil water retention capacity, and meteorological conditions for each irrigation unit and set differentiated irrigation thresholds;

[0030] The adjustment submodule is used to increase the irrigation threshold during the peak plant growth period and adjust the irrigation threshold in advance according to meteorological forecast data during the drought and low rainfall period.

[0031] Furthermore, in the method for implementing the above-mentioned urban garden irrigation intelligent monitoring and management system, the method includes the following steps:

[0032] Through sensors and meteorological equipment, soil temperature and humidity data, garden meteorological data, and plant growth detection data in the garden are collected to obtain initial multi-dimensional garden data;

[0033] Extracting correlation features from the initial multi-dimensional garden data using a machine learning algorithm to obtain multi-dimensional correlated garden data;

[0034] A Bi-LSTM garden dynamic water demand prediction model was established based on the Bi-LSTM neural network. The hyperparameters of the prediction model were optimized using the PSO particle swarm optimization algorithm to obtain the target Bi-LSTM garden dynamic water demand prediction model.

[0035] Inputting the multi-dimensional associated garden data into the target Bi-LSTM garden dynamic water demand prediction model for training to obtain the garden dynamic water demand;

[0036] The garden is divided into multiple irrigation units based on the dynamic water demand of the garden, and differentiated irrigation thresholds are set according to the irrigation units. The water-fertilizer ratio is optimized through the SVM support vector machine algorithm to obtain a garden zoning irrigation strategy, and the garden is irrigated based on the garden zoning irrigation strategy.

[0037] Furthermore, in the method for implementing the above-mentioned urban garden irrigation intelligent monitoring and management system, the method includes the following steps:

[0038] The temperature and humidity data in the soil are collected using a temperature and humidity sensor to obtain soil temperature and humidity data;

[0039] Use anemometers, rain gauges, temperature and humidity sensors, and light intensity sensors to monitor garden meteorological data in real time, including at least wind speed, rainfall, temperature, humidity, image data, and light intensity;

[0040] For different types of garden plants, corresponding growth monitoring sensors are set up, including at least leaf moisture sensors and stem flow sensors, to collect plant growth detection data;

[0041] The Kalman filter algorithm is used to synchronize timestamps and match spatial coordinates of multi-source data collected by sensors, eliminating sensor drift errors and obtaining multi-dimensional garden data.

[0042] Outliers in the multi-dimensional garden data are identified based on the Isolation Forest algorithm, and missing data are filled using the neighboring sensor interpolation method to obtain initial multi-dimensional garden data.

[0043] Furthermore, in the method for implementing the above-mentioned urban garden irrigation intelligent monitoring and management system, the method includes the following steps:

[0044] Calculating the soil moisture change rate, evaporation and rainfall cumulative difference in the initial multi-dimensional garden data through a sliding window to obtain the impact of short-term environmental fluctuations on water demand and obtain time series feature data;

[0045] The CNN convolutional neural network is used to perform semantic segmentation on the image data, extract the spatial distribution characteristics of vegetation density, canopy coverage area and terrain slope, and obtain spatial feature data;

[0046] Based on random forest, the weight of the interaction between soil salinity and pH value on plant water absorption efficiency was calculated, the coupling characteristics of soil and plants were established, and cross-feature data were obtained;

[0047] Based on the GAN generative adversarial network, the soil moisture change curve under different meteorological conditions is simulated, the small sample scene data is expanded, and the extracted feature data is fused to obtain multi-dimensional correlated garden data.

[0048] Its beneficial effects are as follows: soil temperature and humidity data, garden meteorological data, and plant growth detection data in the garden are collected through sensors and meteorological equipment to obtain initial multi-dimensional garden data; correlation features in the initial multi-dimensional garden data are extracted using a machine learning algorithm to obtain multi-dimensional correlated garden data; a Bi-LSTM garden dynamic water demand prediction model is established based on a Bi-LSTM neural network, and the hyperparameters of the prediction model are optimized using a PSO particle swarm optimization algorithm to obtain a target Bi-LSTM garden dynamic water demand prediction model; the multi-dimensional correlated garden data is input into the target Bi-LSTM garden dynamic water demand prediction model for training to obtain the garden dynamic water demand; the garden is divided into multiple irrigation units based on the garden dynamic water demand, and differentiated irrigation thresholds are set according to the irrigation units. The water-fertilizer ratio is optimized using an SVM support vector machine algorithm to obtain a garden zoning irrigation strategy, and the garden is irrigated based on the garden zoning irrigation strategy. 1. It can accurately predict the future water demand of the garden. Managers can plan irrigation arrangements in advance based on the prediction results to avoid the growth of plants affected by water shortage or excessive water, ensure the healthy growth of garden plants, and improve the garden landscape effect. 2. Precise fertilization based on plant nutritional needs improves fertilizer utilization. This reduces garden irrigation and maintenance costs while minimizing fertilizer pollution, achieving efficient utilization of water and fertilizer resources and promoting the sustainable development of green urban gardens. 3. Managers can intuitively access garden information and issue instructions through the system, eliminating the need for frequent on-site inspections. This significantly improves garden management efficiency, reduces labor costs, and makes garden management more refined and standardized. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0050] Figure 1 This is a schematic diagram of a first embodiment of an intelligent monitoring and management system for urban garden irrigation according to an embodiment of the present invention;

[0051] Figure 2 Schematic diagram of a second embodiment of the urban garden irrigation intelligent monitoring and management system according to an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of a third embodiment of the urban garden irrigation intelligent monitoring and management system in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0054] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0055] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, the urban garden irrigation intelligent monitoring and management system includes:

[0056] 101. A garden data acquisition module is used to collect soil temperature and humidity data, garden meteorological data, and plant growth detection data in the garden through sensors and meteorological equipment to obtain initial multi-dimensional garden data;

[0057] Specifically, this embodiment also includes: a soil temperature and humidity data acquisition submodule, which is used to collect temperature data and humidity data in the soil using a temperature and humidity sensor to obtain soil temperature and humidity data;

[0058] The garden meteorological data collection submodule is used to monitor the garden meteorological data in real time using an anemometer, a rain gauge, a temperature and humidity sensor, and a light intensity sensor, including at least wind speed, rainfall, temperature, humidity, image data, and light intensity;

[0059] The plant growth detection data collection submodule is used to set corresponding growth monitoring sensors for different types of garden plants, including at least leaf moisture sensors and stem flow sensors, to collect plant growth detection data;

[0060] The data matching submodule is used to synchronize timestamps and match spatial coordinates of multi-source data collected by sensors using the Kalman filter algorithm, eliminating sensor drift errors and obtaining multi-dimensional garden data.

[0061] The data processing submodule is used to identify outliers in multi-dimensional garden data based on the Isolation Forest algorithm, and use the neighboring sensor interpolation method to fill in missing data to obtain initial multi-dimensional garden data.

[0062] 102. A data processing and analysis module, configured to extract correlation features from the initial multi-dimensional garden data using a machine learning algorithm to obtain multi-dimensional correlated garden data;

[0063] Specifically, this embodiment also includes: a time series feature extraction submodule for calculating the soil moisture change rate, the cumulative difference between evaporation and rainfall in the initial multi-dimensional garden data through a sliding window, obtaining the impact of short-term environmental fluctuations on water demand, and obtaining time series feature data;

[0064] The spatial feature extraction submodule is used to perform semantic segmentation on image data using the CNN convolutional neural network, extract the spatial distribution characteristics of vegetation density, canopy coverage area and terrain slope, and obtain spatial feature data;

[0065] The cross-feature extraction submodule is used to calculate the influence weight of the interaction between soil salinity and pH value on plant water absorption efficiency based on random forest, establish soil and plant coupling characteristics, and obtain cross-feature data;

[0066] The feature data enhancement submodule is used to simulate the soil moisture change curve under different meteorological conditions based on the GAN generative adversarial network, expand the small sample scene data, fuse the extracted feature data, and obtain multi-dimensional correlated garden data.

[0067] 103. A prediction model building module is used to establish a Bi-LSTM garden dynamic water demand prediction model based on a Bi-LSTM neural network, optimize the hyperparameters of the prediction model using a PSO particle swarm optimization algorithm, and obtain a target Bi-LSTM garden dynamic water demand prediction model;

[0068] Specifically, this embodiment also includes: establishing a Bi-LSTM garden dynamic water demand prediction model based on the Bi-LSTM neural network. The Bi-LSTM network synchronously processes time series data through forward and backward LSTM layers to capture irrigation demand and historical meteorological events; the backward layer of the prediction model traces back the soil moisture attenuation trend over the past three days, and the forward layer predicts the transpiration rate based on the light intensity in the next six hours.

[0069] The optimization submodule is used to optimize the hyperparameters of the Bi-LSTM garden dynamic water demand prediction model using the particle swarm optimization (PSO) algorithm;

[0070] The optimization submodule is used to simulate the foraging behavior of bird flocks based on the PSO particle swarm algorithm. It searches for the optimal hyperparameter combination through continuous iterative search of particles in the solution space, including at least the number of network layers, number of neurons, and learning rate.

[0071] The submodule is obtained, which is used to adjust the hyperparameters with the goal of minimizing the prediction error during the optimization process, and obtain the target Bi-LSTM garden dynamic water demand prediction model.

[0072] 104. A water demand prediction module, configured to input multi-dimensional correlated garden data into a target Bi-LSTM garden dynamic water demand prediction model for training to obtain the garden dynamic water demand;

[0073] Specifically, this embodiment also includes: inputting the training set of multi-dimensional correlated garden data into the target Bi-LSTM garden dynamic water demand prediction model for multiple rounds of training. During the training process, the model continuously adjusts internal parameters according to the prediction error and learns the laws and patterns in the data. After each round of training, the model is evaluated using the validation set to observe the prediction performance of the model. When the prediction error of the model on the validation set reaches a satisfactory level or tends to be stable, the training is stopped. Then, the garden data collected in real time (after the same data preprocessing steps as the training set) is input into the trained model, and the model outputs the garden dynamic water demand prediction results to provide a basis for subsequent irrigation decisions.

[0074] 105. The garden zoning irrigation module is used to divide the garden into multiple irrigation units based on the garden's dynamic water demand, set differentiated irrigation thresholds according to the irrigation units, optimize the water-fertilizer ratio through the SVM support vector machine algorithm, obtain the garden zoning irrigation strategy, and irrigate the garden based on the garden zoning irrigation strategy.

[0075] Specifically, this embodiment also includes: dividing the garden into multiple irrigation units based on the garden's topography, plant species distribution, soil type factors, and combined with the garden's dynamic water demand, and dividing the water-loving plant area and the drought-resistant plant area into different irrigation units.

[0076] The threshold submodule is used to obtain the plant growth stage, soil water retention capacity, and meteorological conditions for each irrigation unit and set differentiated irrigation thresholds;

[0077] The adjustment submodule is used to increase the irrigation threshold during the peak plant growth period and adjust the irrigation threshold in advance according to meteorological forecast data during the drought and low rainfall period.

[0078] Its beneficial effects include collecting soil temperature and humidity data, garden meteorological data, and plant growth monitoring data from the garden through sensors and meteorological equipment to obtain initial multi-dimensional garden data; using machine learning algorithms to extract correlation features from the initial multi-dimensional garden data to obtain multi-dimensional correlated garden data; establishing a Bi-LSTM garden dynamic water demand prediction model based on a Bi-LSTM neural network, and optimizing the prediction model's hyperparameters using the Particle Swarm Optimization (PSO) algorithm to obtain a target Bi-LSTM garden dynamic water demand prediction model; inputting the multi-dimensional correlated garden data into the target Bi-LSTM garden dynamic water demand prediction model for training to obtain the garden's dynamic water demand; dividing the garden into multiple irrigation units based on the garden's dynamic water demand, setting differentiated irrigation thresholds for each irrigation unit, and optimizing the water-fertilizer ratio using the Support Vector Machine (SVM) algorithm to obtain a zoned irrigation strategy for the garden, which is then used to irrigate the garden. 1. It accurately predicts the garden's future water demand. Managers can use the predicted results to plan irrigation schedules in advance, preventing plant growth from being affected by water shortages or excesses, ensuring healthy plant growth, and improving the garden's landscape. 2. It precisely applies fertilizers based on plant nutritional needs, improving fertilizer utilization. This not only reduces garden irrigation and maintenance costs, but also minimizes fertilizer pollution, achieving efficient utilization of water and fertilizer resources and promoting the sustainable development of urban gardens. 3. Managers can intuitively access garden information and issue instructions through the system, eliminating the need for frequent on-site inspections. This significantly improves garden management efficiency, reduces labor costs, and makes garden management more refined and standardized.

[0079] See also Figure 2 In a DMTO unit catalyst balance monitoring system based on rapid calculation, the multi-data processing and analysis module includes:

[0080] The time series feature extraction submodule is used to calculate the soil moisture change rate, the cumulative difference between evaporation and rainfall in the initial multi-dimensional garden data through a sliding window, obtain the impact of short-term environmental fluctuations on water demand, and obtain time series feature data;

[0081] The spatial feature extraction submodule is used to perform semantic segmentation on image data using the CNN convolutional neural network, extract the spatial distribution characteristics of vegetation density, canopy coverage area and terrain slope, and obtain spatial feature data;

[0082] The cross-feature extraction submodule is used to calculate the influence weight of the interaction between soil salinity and pH value on plant water absorption efficiency based on random forest, establish soil and plant coupling characteristics, and obtain cross-feature data;

[0083] The feature data enhancement submodule is used to simulate the soil moisture change curve under different meteorological conditions based on the GAN generative adversarial network, expand the small sample scene data, fuse the extracted feature data, and obtain multi-dimensional correlated garden data.

[0084] See also Figure 3 In a DMTO unit catalyst balance monitoring system based on rapid calculation, the prediction model building module includes:

[0085] The optimization submodule is used to optimize the hyperparameters of the Bi-LSTM garden dynamic water demand prediction model using the particle swarm optimization (PSO) algorithm;

[0086] The optimization submodule is used to simulate the foraging behavior of bird flocks based on the PSO particle swarm algorithm. It searches for the optimal hyperparameter combination through continuous iterative search of particles in the solution space, including at least the number of network layers, number of neurons, and learning rate.

[0087] The submodule is obtained, which is used to adjust the hyperparameters with the goal of minimizing the prediction error during the optimization process, and obtain the target Bi-LSTM garden dynamic water demand prediction model.

[0088] Specifically, the embodiment of the present invention can also be implemented in the following ways:

[0089] 1. Data Collection

[0090] (1) Equipment selection and layout planning:

[0091] Before data collection begins, conduct a field survey of the garden and plan the layout of sensors and meteorological equipment based on the garden's area, topography, and plant distribution. Select high-precision and stable soil temperature and humidity sensors, meteorological monitoring equipment (such as a weather station that monitors temperature, humidity, wind speed, light intensity, and rainfall), image recognition cameras, and plant growth sensors. For large gardens, adopt a grid layout with a density of one data collection point every 50-100 square meters to ensure data collection covers the entire garden area.

[0092] (2) Installation and debugging:

[0093] The soil temperature and humidity sensors are buried in the soil surface layer (5-10 cm), middle layer (20-30 cm), and deep layer (40-50 cm) respectively, ensuring that the sensor probe is in full contact with the soil to obtain accurate data. Meteorological equipment is installed in a ventilated, open and unobstructed location to avoid interference with meteorological data collection from buildings, trees, etc. The image recognition camera is installed at a height and angle that can clearly capture the growth status of the plant, and the plant growth sensor is installed in key parts of the plant (such as stems and leaves). After installation, the equipment is debugged and the data transmission is checked to ensure that the collected data is accurate and stable.

[0094] (3) Data collection and storage:

[0095] Once the equipment is operating normally, it automatically collects soil temperature and humidity data, garden meteorological data, and plant growth detection data at a set frequency (e.g., every 15 minutes). Soil temperature and humidity data records the real-time temperature and humidity values ​​of soil layers at different depths; meteorological data records the instantaneous values ​​of various meteorological indicators; and for plant growth detection data, an image recognition camera captures plant images, while plant growth sensors record physiological indicators such as changes in plant stem diameter and leaf moisture content. The collected data is transmitted via a wireless network to a data storage server for classification and storage, forming the initial multi-dimensional garden data.

[0096] 2. Correlation feature extraction

[0097] (1) Data preprocessing:

[0098] The stored initial multi-dimensional garden data was imported into the data processing platform for data cleaning, removing outliers, duplicates, and missing values. Missing values ​​were filled using methods such as mean filling and interpolation. The data was then standardized to bring data of different dimensions to the same scale for ease of subsequent analysis.

[0099] (2) Application of Random Forest Algorithm:

[0100] The preprocessed data is processed using the random forest algorithm. First, the parameters of the random forest are determined, such as the number of decision trees (usually set to 50-200 trees), the feature selection ratio, etc. When the algorithm is running, multiple sample subsets are randomly extracted from the original data with replacement to provide training data for each decision tree. During the training process, each decision tree searches for the optimal partitioning features from the randomly selected feature subsets to construct a decision tree model. Through the parallel training and result integration of multiple decision trees, the potential correlation features between soil temperature and humidity, meteorological conditions and plant growth indicators are explored. For example, the specific correlation degree between soil moisture and plant growth rate within a certain temperature and light intensity range is analyzed, and the key features that have a significant impact on garden water demand are screened out to obtain multi-dimensional correlated garden data.

[0101] 3. Construction and optimization of Bi-LSTM garden dynamic water demand prediction model,

[0102] (1) Model construction preparation:

[0103] Before building a Bi-LSTM model for predicting dynamic garden water demand, it is necessary to organize the multi-dimensional, correlated garden data into time series, arrange the data in chronological order, and divide it into training, validation, and test sets (generally in a ratio of 7:1:2). At the same time, the model's input features must be determined, and features closely related to garden water demand, such as soil temperature and humidity at different depths, air temperature, light intensity, and plant growth indicators, must be selected as model inputs.

[0104] (2) Bi-LSTM model construction:

[0105] Based on a deep learning framework, a Bi-LSTM neural network model was constructed. The number of model layers (generally 2-3) and the number of neurons in each layer were determined (adjusted based on data complexity and model performance). The training data was input into the model. The Bi-LSTM neural network processed the data bidirectionally, learning both forward and backward information, capturing long-term temporal dependencies of the data and predicting the future dynamic water demand of the garden. During training, an appropriate loss function (such as the mean squared error loss function) and optimizer (such as the Adam optimizer) were used to continuously adjust the model parameters to minimize the error between the model's predicted value and the actual water demand.

[0106] (3) PSO algorithm optimizes hyperparameters:

[0107] The particle swarm optimization (PSO) algorithm is used to optimize the hyperparameters of the Bi-LSTM model. First, the hyperparameters that need to be optimized, such as the number of network layers, the number of neurons, and the learning rate, are determined, and the range of the hyperparameter values ​​is set. The PSO algorithm initializes a group of particles, each of which represents a set of hyperparameter combinations. The particles continuously update their positions in the solution space based on their own historical optimal positions and the historical optimal positions of the group, searching for the optimal hyperparameter combination. In each iteration, the current hyperparameter combination is applied to the Bi-LSTM model for training and verification, and the prediction error on the verification set is used as the fitness function to evaluate the quality of the hyperparameter combination. After multiple iterations, when the fitness function value no longer decreases significantly or reaches the preset number of iterations, the optimal hyperparameter combination is obtained and applied to the Bi-LSTM model to obtain the target Bi-LSTM garden dynamic water demand prediction model.

[0108] 4. Model training and water demand prediction

[0109] The training set of multi-dimensional correlated garden data is fed into the target Bi-LSTM garden dynamic water demand prediction model for multiple rounds of training. During training, the model continuously adjusts its internal parameters based on prediction errors, learning patterns and regularities in the data. After each round of training, the model is evaluated using a validation set to observe its predictive performance. Training is terminated when the prediction error on the validation set reaches a satisfactory level or stabilizes. Next, real-time garden data (which has undergone the same data preprocessing steps as the training set) is fed into the trained model, which outputs a garden dynamic water demand prediction result, providing a basis for subsequent irrigation decisions.

[0110] 5. Irrigation strategy formulation

[0111] (1) Irrigation unit division:

[0112] Combined with the prediction results of the garden's dynamic water demand, conduct a field survey of the garden. Divide the garden into multiple irrigation units based on factors such as the garden's topography (such as slope and terrain height), plant species distribution (water-loving plant area, drought-tolerant plant area, flower area, tree area, etc.), and soil type (sandy soil, clay, loam, etc.). During the division process, ensure that the plants in each irrigation unit have similar water demand characteristics, and that the terrain and soil conditions are relatively consistent to facilitate unified management and irrigation. For example, an area with flat terrain, similar soil water retention capacity, and the same type of plants planted can be divided into one irrigation unit.

[0113] (2) Irrigation threshold setting:

[0114] For each irrigation unit, differentiated irrigation thresholds are set based on a comprehensive consideration of factors such as the plant's growth stage (seedling, growing, flowering, dormant, etc.), soil water retention capacity, and meteorological conditions (recent weather forecasts and historical meteorological data for the same period). For seedling plants, due to their weaker root systems, the irrigation threshold can be set lower to maintain moderate soil moisture. During periods of vigorous growth, water demand increases, and the irrigation threshold can be appropriately raised. During periods of drought and low rainfall, the irrigation threshold is lowered in advance based on meteorological forecasts to ensure plant water needs. At the same time, historical data for the same period is referenced to analyze plant water requirements under different meteorological conditions, allowing for dynamic adjustments to the irrigation threshold.

[0115] (3) Optimization of water-fertilizer ratio:

[0116] The SVM support vector machine algorithm is used to optimize the water-fertilizer ratio for each irrigation unit. First, water demand data, plant nutrient requirement data (determined by plant species and growth stage), and soil nutrient data (obtained through soil testing) for different irrigation units are collected. This data is used as input features for the SVM algorithm, and the growth effects of plants under different water-fertilizer ratios (such as plant height, leaf color, flowering and fruiting) are used as output labels to train the SVM model. The SVM algorithm finds the optimal combination of fertilizer concentration and irrigation water volume by searching for the optimal classification hyperplane in high-dimensional space, thereby generating a scientific and reasonable garden zoning irrigation strategy. Based on this strategy, the garden management department controls the irrigation equipment and fertilization equipment in each irrigation unit to achieve precise irrigation and fertilization.

[0117] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention, and all such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. Urban garden irrigation intelligent monitoring and management system, characterized by: The urban garden irrigation intelligent monitoring and management system includes: The garden data acquisition module is used to collect soil temperature and humidity data, garden meteorological data, and plant growth detection data in the garden through sensors and meteorological equipment to obtain initial multi-dimensional garden data; A data processing and analysis module is used to extract correlation features from the initial multi-dimensional garden data using a machine learning algorithm to obtain multi-dimensional correlated garden data; The prediction model building module is used to establish a Bi-LSTM garden dynamic water demand prediction model based on the Bi-LSTM neural network, and use the PSO particle swarm optimization algorithm to optimize the hyperparameters of the prediction model to obtain the target Bi-LSTM garden dynamic water demand prediction model; A water demand prediction module is used to input the multi-dimensional associated garden data into the target Bi-LSTM garden dynamic water demand prediction model for training to obtain the garden dynamic water demand; The garden zoning irrigation module is used to divide the garden into multiple irrigation units based on the dynamic water demand of the garden, set differentiated irrigation thresholds according to the irrigation units, optimize the water-fertilizer ratio through the SVM support vector machine algorithm, obtain the garden zoning irrigation strategy, and irrigate the garden based on the garden zoning irrigation strategy.

2. The urban garden irrigation intelligent monitoring and management system according to claim 1, characterized in that: The garden data acquisition module includes the following submodules: The soil temperature and humidity data acquisition submodule is used to collect temperature and humidity data in the soil using temperature and humidity sensors to obtain soil temperature and humidity data; The garden meteorological data collection submodule is used to monitor the garden meteorological data in real time using an anemometer, a rain gauge, a temperature and humidity sensor, and a light intensity sensor, including at least wind speed, rainfall, temperature, humidity, image data, and light intensity; The plant growth detection data collection submodule is used to set corresponding growth monitoring sensors for different types of garden plants, including at least leaf moisture sensors and stem flow sensors, to collect plant growth detection data; The data matching submodule is used to synchronize timestamps and match spatial coordinates of multi-source data collected by sensors using the Kalman filter algorithm, eliminating sensor drift errors and obtaining multi-dimensional garden data. The data processing submodule is used to identify outliers in the multi-dimensional garden data based on the Isolation Forest algorithm, and fill in missing data using the neighboring sensor interpolation method to obtain initial multi-dimensional garden data.

3. The urban garden irrigation intelligent monitoring and management system according to claim 1, characterized in that: The data processing and analysis module includes the following submodules: A time series feature extraction submodule is used to calculate the soil moisture change rate, the cumulative difference between evaporation and rainfall in the initial multi-dimensional garden data through a sliding window, obtain the impact of short-term environmental fluctuations on water demand, and obtain time series feature data; The spatial feature extraction submodule is used to perform semantic segmentation on image data using the CNN convolutional neural network, extract the spatial distribution characteristics of vegetation density, canopy coverage area and terrain slope, and obtain spatial feature data; The cross-feature extraction submodule is used to calculate the influence weight of the interaction between soil salinity and pH value on plant water absorption efficiency based on random forest, establish soil and plant coupling characteristics, and obtain cross-feature data; The feature data enhancement submodule is used to simulate the soil moisture change curve under different meteorological conditions based on the GAN generative adversarial network, expand the small sample scene data, fuse the extracted feature data, and obtain multi-dimensional correlated garden data.

4. The urban garden irrigation intelligent monitoring and management system according to claim 1, characterized in that: The prediction model establishment module includes: establishing a Bi-LSTM garden dynamic water demand prediction model based on a Bi-LSTM neural network. The Bi-LSTM network synchronously processes time series data through forward and backward LSTM layers to capture irrigation demand and historical meteorological events; the backward layer of the prediction model traces the soil moisture attenuation trend over the past three days, and the forward layer predicts the transpiration rate based on the light intensity in the next six hours.

5. The urban garden irrigation intelligent monitoring and management system according to claim 1, characterized in that: The prediction model building module also includes the following submodules: The optimization submodule is used to optimize the hyperparameters of the Bi-LSTM garden dynamic water demand prediction model using the particle swarm optimization (PSO) algorithm; The optimization submodule is used to simulate the foraging behavior of bird flocks based on the PSO particle swarm algorithm. It searches for the optimal hyperparameter combination through continuous iterative search of particles in the solution space, including at least the number of network layers, number of neurons, and learning rate. The submodule is obtained, which is used to adjust the hyperparameters with the goal of minimizing the prediction error during the optimization process, and obtain the target Bi-LSTM garden dynamic water demand prediction model.

6. The urban garden irrigation intelligent monitoring and management system according to claim 1, characterized in that: The garden zoning irrigation module includes: dividing the garden into multiple irrigation units according to the garden's topography, plant species distribution, soil type factors, and combined with the garden's dynamic water demand, and dividing the water-loving plant area and the drought-resistant plant area into different irrigation units.

7. The urban garden irrigation intelligent monitoring and management system according to claim 1, characterized in that: The garden partition irrigation module includes the following submodules: The threshold submodule is used to obtain the plant growth stage, soil water retention capacity, and meteorological conditions for each irrigation unit and set differentiated irrigation thresholds; The adjustment submodule is used to increase the irrigation threshold during the peak plant growth period and adjust the irrigation threshold in advance according to meteorological forecast data during the drought and low rainfall period.

8. The method for implementing the urban garden irrigation intelligent monitoring and management system as claimed in claim 1 is characterized in that: The method comprises the following steps: Through sensors and meteorological equipment, soil temperature and humidity data, garden meteorological data, and plant growth detection data in the garden are collected to obtain initial multi-dimensional garden data; Extracting correlation features from the initial multi-dimensional garden data using a machine learning algorithm to obtain multi-dimensional correlated garden data; A Bi-LSTM garden dynamic water demand prediction model was established based on the Bi-LSTM neural network. The hyperparameters of the prediction model were optimized using the PSO particle swarm optimization algorithm to obtain the target Bi-LSTM garden dynamic water demand prediction model. Inputting the multi-dimensional associated garden data into the target Bi-LSTM garden dynamic water demand prediction model for training to obtain the garden dynamic water demand; The garden is divided into multiple irrigation units based on the dynamic water demand of the garden, and differentiated irrigation thresholds are set according to the irrigation units. The water-fertilizer ratio is optimized through the SVM support vector machine algorithm to obtain a garden zoning irrigation strategy, and the garden is irrigated based on the garden zoning irrigation strategy.

9. The method for implementing the urban garden irrigation intelligent monitoring and management system as claimed in claim 1 is characterized in that: The method comprises the following steps: The temperature and humidity data in the soil are collected using a temperature and humidity sensor to obtain soil temperature and humidity data; Use anemometers, rain gauges, temperature and humidity sensors, and light intensity sensors to monitor garden meteorological data in real time, including at least wind speed, rainfall, temperature, humidity, image data, and light intensity; For different types of garden plants, corresponding growth monitoring sensors are set up, including at least leaf moisture sensors and stem flow sensors, to collect plant growth detection data; The Kalman filter algorithm is used to synchronize timestamps and match spatial coordinates of multi-source data collected by sensors, eliminating sensor drift errors and obtaining multi-dimensional garden data. Outliers in the multi-dimensional garden data are identified based on the Isolation Forest algorithm, and missing data are filled using the neighboring sensor interpolation method to obtain initial multi-dimensional garden data.

10. A method for implementing the urban garden irrigation intelligent monitoring and management system as claimed in claim 1, characterized in that: The method comprises the following steps: Calculating the soil moisture change rate, evaporation and rainfall cumulative difference in the initial multi-dimensional garden data through a sliding window to obtain the impact of short-term environmental fluctuations on water demand and obtain time series feature data; The CNN convolutional neural network is used to perform semantic segmentation on the image data, extract the spatial distribution characteristics of vegetation density, canopy coverage area and terrain slope, and obtain spatial feature data; Based on random forest, the weight of the interaction between soil salinity and pH value on plant water absorption efficiency was calculated, the coupling characteristics of soil and plants were established, and cross-feature data were obtained; Based on the GAN generative adversarial network, the soil moisture change curve under different meteorological conditions is simulated, the small sample scene data is expanded, and the extracted feature data is fused to obtain multi-dimensional correlated garden data.

Images