A smart soil moisture content analysis system based on an internet of things terminal

Abstract

The application relates to the field of agricultural information technology, and particularly discloses a smart soil moisture content analysis system based on an Internet of Things terminal. The system comprises an Internet of Things terminal network composed of heterogeneous nodes integrated with multiple types of soil moisture sensors, an edge computing gateway, and a cloud analysis platform. The energy collection and management module ensures long-term operation of the nodes, the edge gateway performs data cleaning and preliminary interpolation, the cloud platform uses a space-time collaborative Kriging interpolation engine to fuse discrete profile point data and continuous surface trend data, construct a high-precision three-dimensional continuous soil moisture content body data model, and support water migration simulation and irrigation decision-making, thereby realizing accurate, low-cost and sustainable monitoring and analysis of the three-dimensional distribution of farmland soil moisture content.

Description

Technical Field

[0001] This invention belongs to the field of agricultural information technology, specifically relating to a smart soil moisture analysis system based on an Internet of Things (IoT) terminal. Background Technology

[0002] In the field of smart agriculture, precise sensing and regulation of the crop growth environment are key to achieving efficient resource utilization and yield improvement. Soil moisture, as a core environmental factor affecting crop growth, requires precise, continuous, and large-scale monitoring and analysis. This is fundamental to implementing precision irrigation decisions and is of great significance for ensuring food security and sustainable agricultural development.

[0003] Soil moisture monitoring and analysis is an important component of smart agriculture technology systems. This technology aims to acquire soil moisture information through sensor networks and use data analysis methods to assess soil moisture conditions and their spatiotemporal variations, thereby providing a scientific basis for agricultural production management.

[0004] Current technologies primarily rely on deploying fixed or mobile discrete point sensors to collect soil moisture data. This point-based monitoring method has significant limitations: First, to obtain representative field soil moisture information, a large number of sensors need to be densely deployed in vast farmlands, resulting in high hardware costs and complex deployment and maintenance. Second, discrete point data cannot effectively depict the true distribution of soil moisture in continuous three-dimensional space in the field, and cannot generate high-precision soil moisture distribution maps, thus limiting the effectiveness of precision irrigation. Furthermore, traditional sensors typically require continuous external power or frequent battery replacements, making it difficult to ensure power supply in field environments, which affects the long-term stable operation of the monitoring system.

[0005] Therefore, how to achieve continuous, three-dimensional, and large-scale accurate perception and analysis of farmland soil moisture in a low-cost and low-power manner has become a technical problem that urgently needs to be solved in the field of smart agriculture. Summary of the Invention

[0006] This invention provides a smart soil moisture analysis system based on an Internet of Things (IoT) terminal to solve the technical contradictions in existing technologies, such as high cost due to reliance on discrete point sensors, inability to depict continuous three-dimensional soil moisture distribution, and difficulty in providing power in the field.

[0007] To achieve the above objectives, this invention provides a smart soil moisture analysis system based on Internet of Things (IoT) terminals. The system includes an IoT terminal network deployed in a target farmland area, an edge computing gateway, and a cloud-based analysis platform. The IoT terminal network consists of multiple heterogeneous sensor nodes, each integrating at least two different types of soil moisture sensors, a microcontroller unit, a low-power wide-area network communication module, and an energy harvesting and management module. The edge computing gateway is deployed at the edge of the farmland area to aggregate and preprocess data from the IoT terminal network. The cloud-based analysis platform is used to perform deep data fusion and three-dimensional soil moisture modeling.

[0008] Furthermore, the heterogeneous sensing node includes at least two different types of soil moisture sensors, specifically a profile moisture sensor based on the frequency domain reflection principle and a surface moisture sensor based on the thermal pulse principle. The profile moisture sensor employs a multi-probe array structure, with its probes vertically inserted into the soil at non-equidistant intervals, enabling simultaneous measurement of soil volumetric water content at depths of 5 cm, 15 cm, 30 cm, and 50 cm below the surface. The surface moisture sensor is installed close to the surface and is used to measure soil moisture conditions within a depth range of 0 to 2 cm below the surface. The microcontroller unit periodically and synchronously acquires raw voltage signals from the two sensors at a preset sampling frequency and converts the voltage signals into standardized soil volumetric water content values ​​using a built-in conversion algorithm pre-calibrated for each sensor model.

[0009] Furthermore, the energy harvesting and management module includes a solar photovoltaic panel, a supercapacitor bank, and a power management integrated circuit. The solar photovoltaic panel provides the primary energy source for the entire heterogeneous sensing node. The power management integrated circuit monitors the output power of the solar photovoltaic panel and the remaining charge of the supercapacitor bank in real time. When the charge of the supercapacitor bank falls below a first preset threshold, the power management integrated circuit switches the system to an ultra-low power sleep mode, maintaining only the basic clock and wake-up timer, and stopping all sensor data acquisition and wireless communication. When the output power of the solar photovoltaic panel continuously exceeds a second preset threshold and the charge of the supercapacitor bank recovers to above a third preset threshold, the power management integrated circuit controls the system to exit sleep mode and resume normal operation. This design ensures that the system can maintain continuous operation for more than 30 days through intermittent operation during continuous rainy weather.

[0010] Furthermore, the low-power wide-area network (LPWAN) communication module employs a communication protocol based on spread spectrum technology. Upon each wake-up, the microcontroller unit of each heterogeneous sensor node first encapsulates the multi-depth soil moisture content data collected and converted within the current cycle, along with the node's own GPS positioning coordinates and acquisition timestamp, into a data packet. Subsequently, the LPWAN communication module transmits this data packet to the edge computing gateway at a transmit power of less than 20 milliwatts. Data packet transmission employs a random backoff and acknowledgment retransmission mechanism to ensure communication reliability under dense node deployment.

[0011] Furthermore, the edge computing gateway includes a multi-protocol access unit, a data cleaning unit, and a preliminary spatial interpolation unit. The multi-protocol access unit is responsible for receiving data packets from all heterogeneous sensor nodes and parsing the location coordinates of each node and its corresponding multi-depth soil moisture content dataset. The data cleaning unit performs outlier detection and removal on the received dataset. Specifically, for moisture content data at the same depth layer, it calculates the average and standard deviation of all node data; single data points whose values ​​deviate from the average by more than twice the standard deviation are marked as outliers; for data points marked as outliers, the linear interpolation result of the two adjacent valid historical measurements of that data point is used to replace it. The preliminary spatial interpolation unit processes the cleaned surface soil moisture content data from the same sampling time, using an inverse distance weighted interpolation algorithm to generate a two-dimensional surface soil moisture distribution raster map covering the entire farmland area with a resolution of 1 meter × 1 meter.

[0012] Furthermore, the cloud-based analysis platform includes a data receiving and storage service, a spatiotemporal co-kriging interpolation engine, and a three-dimensional soil moisture model construction module. The data receiving and storage service continuously receives two-dimensional surface soil moisture distribution raster maps from the edge computing gateway and raw multi-depth profile data from all nodes, storing them in chronological order. The spatiotemporal co-kriging interpolation engine is the core analysis component of the system, and its operation process is as follows: First, the engine extracts the two-dimensional surface soil moisture distribution raster data for the current moment and the previous four consecutive moments from the storage service, using it as spatially continuous trend surface data. Simultaneously, it extracts the profile depth data collected by all heterogeneous sensing nodes within the same time period, using it as spatially discrete verification point data with vertical dimensions. Second, the engine constructs a composite covariance function model that integrates spatial autocorrelation, temporal autocorrelation, and spatial depth correlation. This composite covariance function model collaboratively correlates the surface trend surface data and profile point data in the spatiotemporal depth three-dimensional domain. Finally, the engine performs co-kriging interpolation calculations to find the optimal unbiased estimate of soil volumetric moisture content at any horizontal position and any depth for undeployed nodes, and finally outputs a continuously distributed soil moisture data model with three-dimensional spatial coordinates.

[0013] Furthermore, the three-dimensional soil moisture model construction module receives the soil moisture volume data model output by the spatiotemporal co-kriging interpolation engine and converts it into a renderable and analyzable three-dimensional data structure. This soil moisture volume data module can dynamically generate high-precision two-dimensional soil moisture distribution thematic maps based on any horizontal slice depth or vertical profile position specified by the user. Simultaneously, the soil moisture volume data module integrates a water transport simulation algorithm, which can simulate the redistribution of soil moisture over the next 6, 12, and 24 hours based on the three-dimensional soil moisture volume data model, and predict the changing trend of effective water content at different depths in the crop root zone.

[0014] Furthermore, the cloud-based analysis platform is also connected to an irrigation decision interface module. Based on the crop root zone effective water content prediction trend provided by the three-dimensional soil moisture model construction module, and combined with pre-input parameters such as crop type, growth stage, and soil type, the irrigation decision interface module calls upon a built-in irrigation decision model. This model automatically generates a control instruction set containing suggested irrigation time, duration, and amount by comparing the predicted effective water content with the crop's water requirement threshold at that growth stage. This set is then distributed to the intelligent irrigation execution system in the farmland via an application programming interface.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0016] 1. This invention, by designing a heterogeneous sensing node integrating multiple types of sensors and employing a strategy of simultaneous profile and surface sensing, achieves multi-depth continuous sensing of the vertical soil profile on a single node, breaking through the limitations of traditional single-point, single-layer measurements. Combining low-power wide-area networks and intelligent energy management, it significantly reduces sensor deployment density per unit area while using an advanced spatiotemporal co-Kriging interpolation algorithm to deeply fuse sparse, discrete profile point data with continuous surface trend data, successfully constructing a high-precision three-dimensional continuous soil moisture data model. This fundamentally solves the contradiction of relying on dense point-like networks to approximate spatial continuous distribution, achieving accurate depiction of the true three-dimensional distribution of farmland soil moisture at a lower overall cost.

[0017] 2. The energy harvesting and management module constructed in this invention, through the combination of solar photovoltaic panels and supercapacitor banks, and combined with a multi-threshold sleep control strategy based on power and electricity monitoring, achieves autonomous energy acquisition and intelligent allocation. This system can automatically enter a deep sleep state to conserve energy under severe weather conditions and automatically wake up after energy recovery, ensuring long-term, stable, and uninterrupted operation of the monitoring network in field environments without a stable mains power supply. This effectively solves the core pain point of difficult power supply guarantee for traditional monitoring equipment, greatly improving the system's practicality and reliability.

[0018] 3. The system architecture of this invention forms a complete technical closed loop from terminal perception and edge preprocessing to cloud-based deep analysis. The edge computing gateway performs real-time data cleaning and preliminary spatial interpolation, reducing the cloud load and providing a rapid surface soil moisture view. The cloud platform focuses on complex spatiotemporal deep 3D modeling and predictive analysis. Its spatiotemporal co-kriging interpolation engine fully utilizes the spatial continuity of surface data and the vertical accuracy of profile data, resulting in a 3D model with both high resolution and high reliability. Based on this model, the water transport simulation and irrigation decision-making functions elevate soil moisture analysis from static monitoring to dynamic prediction and precise control, providing a direct and scientific basis for implementing truly on-demand precision irrigation and significantly improving the efficiency of agricultural water resource utilization. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention;

[0020] Figure 2 This is a schematic diagram of the core principle framework of the spatiotemporal collaborative kriging interpolation engine in this invention;

[0021] Figure 3 This is a logical flowchart of the heterogeneous sensing node energy harvesting and management module in this invention.

[0022] Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow from heterogeneous sensing nodes to the cloud analysis platform in this invention;

[0023] Figure 5 This is a logical flowchart of the three-dimensional soil moisture model construction and irrigation decision support in this invention. Detailed Implementation

[0024] Example 1: Please refer to the appendix Figure 1 To be continued Figure 5 This invention provides a smart soil moisture analysis system based on Internet of Things (IoT) terminals. The system consists of three main parts: an IoT terminal network deployed in the target farmland area, an edge computing gateway, and a cloud analysis platform, forming a complete closed-loop system from on-site perception and edge preprocessing to deep cloud modeling and decision support. The entire system's operation logic strictly follows a linear process of data acquisition, transmission, cleaning, interpolation, modeling, simulation, and irrigation control. The components are highly coordinated in function and clearly hierarchical in structure.

[0025] First, the IoT terminal network consists of several heterogeneous sensing nodes, which are deployed in a non-uniform but comprehensive manner within the farmland area to be monitored. Each heterogeneous sensing node, as the front-end sensing unit of the system, integrates at least two different types of soil moisture sensors, a microcontroller unit, a low-power wide-area network communication module, and an energy harvesting and management module.

[0026] The two different types of soil moisture sensors specifically include a profile moisture sensor based on the frequency domain reflection principle and a surface moisture sensor based on the thermal pulse principle. The profile moisture sensor employs a multi-probe array structure, with probes inserted vertically into the soil at non-equidistant intervals, enabling simultaneous measurement of soil volumetric water content at depths of 5 cm, 15 cm, 30 cm, and 50 cm below the surface. The surface moisture sensor is installed close to the surface to accurately acquire soil moisture conditions within a depth range of 0 to 2 cm. The microcontroller unit periodically and synchronously acquires raw voltage signals from the two sensors at a preset sampling frequency and converts these signals into standardized soil volumetric water content values ​​using a built-in conversion algorithm pre-calibrated for each sensor model. This synchronous acquisition mechanism ensures the consistency and timeliness of vertical profile and surface data at the same point in time, providing a high-quality input foundation for subsequent 3D modeling.

[0027] Furthermore, the energy harvesting and management module equipped in each heterogeneous sensing node is a key component ensuring the long-term operation of the system in the field. This energy harvesting and management module includes solar photovoltaic panels, supercapacitor banks, and a power management integrated circuit. The solar photovoltaic panels serve as the primary energy source, continuously powering the entire node. The power management integrated circuit monitors the output power of the solar photovoltaic panels and the remaining charge status of the supercapacitor banks in real time. When the charge of the supercapacitor banks is detected to be lower than a first preset threshold (e.g., set at 20% of the total capacity), the power management integrated circuit immediately controls the system to enter an ultra-low power sleep mode. In this mode, only the basic real-time clock and wake-up timer are maintained, while all sensor data acquisition operations and wireless communication activities are stopped, thereby maximizing energy savings.

[0028] When the output power of the solar photovoltaic panel continuously exceeds the second preset threshold (e.g., set to 50 milliwatts) and the charge of the supercapacitor bank recovers to above the third preset threshold (e.g., set to 60% of the total capacity), the power management integrated circuit triggers the system to exit sleep mode and restore the normal operation of all functional modules. (See attached diagram.) Figure 3 The logical flow framework shown demonstrates that this energy management strategy enables the system to maintain continuous operation for more than 30 days even under continuous rainy weather conditions through intermittent working mode, completely solving the maintenance problems caused by traditional soil monitoring equipment relying on mains power or frequent battery replacements.

[0029] After completing local data acquisition and preliminary processing, each heterogeneous sensor node uploads data to the edge computing gateway via its low-power wide-area network (LPWAN) communication module. This LPWAN communication module employs a spread spectrum-based communication protocol, possessing strong anti-interference capabilities and long-distance transmission characteristics. Each time a node is woken up, the microcontroller unit first encapsulates the multi-depth soil moisture content data acquired and converted within the current cycle, along with the node's own GPS positioning coordinates and second-accurate timestamp information, into a structured data packet.

[0030] Subsequently, the low-power wide-area network communication module transmits the data packet to the edge computing gateway located at the edge of the farmland area with a transmit power of less than 20 milliwatts. To address channel contention issues that may arise from high-density node deployment, a random backoff and acknowledgment retransmission mechanism is employed during data packet transmission: if no acknowledgment response is received from the gateway within a specified time, the node will retry transmitting after a random delay window until a successful acknowledgment signal is received. This mechanism effectively improves communication reliability and data integrity in large-scale deployment scenarios.

[0031] As the intermediate processing layer of the system, the edge computing gateway undertakes the core tasks of data aggregation, cleaning, and preliminary spatial interpolation. Its internal structure comprises three functional sub-modules: a multi-protocol access unit, a data cleaning unit, and a preliminary spatial interpolation unit. The multi-protocol access unit is responsible for receiving data packets from all heterogeneous sensor nodes and parsing them according to protocol specifications to obtain the geographical coordinates of each node and its corresponding multi-depth soil moisture content dataset. The data cleaning unit performs outlier detection and correction operations on the received raw dataset.

[0032] The specific process is as follows: For the water content data reported by all nodes at the same depth layer (e.g., the 5 cm layer), first calculate the arithmetic mean of the data at that same depth layer. with standard deviation Then the values ​​deviate from the mean by more than 2 standard deviations (i.e. Individual data points are marked as outliers. For these outliers, the result of linear interpolation between two adjacent valid historical measurements is used to replace them, thus eliminating the impact of outliers caused by sensor malfunctions or sudden environmental changes on the overall analysis results. The preliminary spatial interpolation unit focuses on processing the cleaned surface soil moisture content data at the same sampling time. This unit uses the Inverse Distance Weighting (IDW) algorithm to generate a 1m × 1m resolution two-dimensional surface soil moisture distribution raster map covering the entire farmland area based on the spatial location of each node and its corresponding surface moisture content value. This two-dimensional surface soil moisture distribution raster map not only intuitively reflects the spatial variation characteristics of surface soil moisture but also provides key trend surface constraints for subsequent cloud-based 3D modeling.

[0033] The cloud-based analytics platform, as the highest level of the system, is responsible for performing advanced intelligent analysis tasks such as deep data fusion, 3D soil moisture modeling, water transport simulation, and irrigation decision support. Its core components include a data receiving and storage service, a spatiotemporal co-kriging interpolation engine, and a 3D soil moisture model construction module. The data receiving and storage service continuously receives 2D surface soil moisture distribution raster maps from the edge computing gateway and raw multi-depth profile data uploaded by all heterogeneous sensor nodes, and stores it in a structured manner according to time series, forming a historical database that can be queried and accessed in batches. The spatiotemporal co-kriging interpolation engine is the core technical component for achieving high-precision 3D modeling in this system; its operating principle is described in the attached figure. Figure 2 As shown.

[0034] The spatiotemporal co-Kriging interpolation engine's workflow consists of three stages: the first stage involves extracting the current moment from the storage service. and the first 4 consecutive moments ( , , , Two-dimensional surface soil moisture distribution raster data was used as spatially continuous trend surface data. Simultaneously, profile depth data (including four depth layers: 5 cm, 15 cm, 30 cm, and 50 cm) collected by all heterogeneous sensing nodes within the same time period were extracted and used as spatially discrete validation point data with vertical dimension. In the second stage, a composite covariance function model integrating spatial autocorrelation, temporal autocorrelation, and spatial depth correlation was constructed. This composite covariance function model can be expressed as:

[0035]

[0036] in, Indicates spatial lag distance Time lag interval and depth lag difference The covariance value under the following conditions; For prior variance; , , These are the range parameters in the spatial, temporal, and depth directions, used to characterize the strength of correlation in each dimension. This covariance function collaboratively correlates surface trend data and profile point data in the three-dimensional domain of space, time, and depth, fully exploring the inherent dependencies between the data. In the third stage, based on the above covariance model, co-kriging interpolation is performed to obtain the optimal unbiased estimate of soil volumetric water content at any horizontal position (x, y) and any depth z for undeployed nodes. The final output is a continuously distributed soil moisture data model with three-dimensional spatial coordinates. This soil moisture data model not only retains the vertical accuracy provided by the profile sensor, but also inherits the spatial continuity of the surface grid map, achieving a high degree of approximation to the real three-dimensional distribution of soil moisture.

[0037] The 3D soil moisture model building module receives soil moisture volume data models output by a spatiotemporal co-kriging interpolation engine and converts them into a 3D data structure suitable for visualization rendering and quantitative analysis. This module supports interactive operation and can dynamically generate high-precision 2D soil moisture distribution thematic maps based on any specified horizontal slice depth (e.g., 25 cm deep at the center of the crop root zone) or any vertical profile location (e.g., a longitudinal profile along a field ridge). Furthermore, the module integrates a water transport simulation algorithm, which can simulate the redistribution of soil moisture over the next 6, 12, and 24 hours based on the current 3D soil moisture volume data model, combined with meteorological forecast data (e.g., future precipitation and evaporation) and soil physical parameters (e.g., hydraulic conductivity and water holding capacity). The simulation results are presented in time series form, predicting the trend of effective water content changes at different depths in the crop root zone, providing agricultural managers with a forward-looking decision-making basis.

[0038] To further enhance the system's practical value, the cloud-based analysis platform also includes an irrigation decision interface module. This module, based on the crop root zone effective water content prediction trend provided by the 3D soil moisture model construction module, and combined with pre-inputted crop type (e.g., corn, wheat), growth stage (e.g., jointing stage, grain-filling stage), and soil type parameters (e.g., sandy loam, clay), calls upon the built-in irrigation decision model. The core logic of this model is to compare the predicted effective water content with the crop's water requirement threshold at that growth stage. If the predicted value is lower than the threshold, irrigation is deemed necessary; otherwise, the status quo is maintained. Based on this, the model comprehensively considers factors such as soil infiltration rate, irrigation system efficiency, and water availability, automatically generating a control instruction set including suggested irrigation time, duration, and amount. This control instruction set is distributed to the intelligent irrigation execution system (e.g., electric valves, drip irrigation controllers) in the farmland through a standardized application programming interface, thereby achieving fully automated closed-loop control from soil moisture perception to precise irrigation.

[0039] In summary, the IoT-based smart soil moisture analysis system described in this embodiment constructs an efficient, reliable, and low-cost three-dimensional soil moisture monitoring and control system through the multi-dimensional sensing capabilities of heterogeneous sensor nodes, the real-time preprocessing capabilities of edge computing gateways, and the deep modeling and decision-making capabilities of cloud analysis platforms. This system not only overcomes the shortcomings of traditional point-based sensor networks, such as high deployment costs and inability to depict continuous distribution, but also solves the problem of power supply in the field through intelligent energy management strategies. Ultimately, it achieves precise, dynamic, and comprehensive control over farmland soil moisture status, providing solid technical support for water conservation, efficiency improvement, and sustainable development in modern agriculture.

[0040] Example 2: Based on Example 1, this example optimizes and expands some components of the system to adapt to more complex or special farmland application scenarios, while further improving the system's robustness and adaptability. The overall system architecture of this example remains as shown in the attached figure. Figure 1 As shown, however, new technical details have been introduced in areas such as sensor configuration of heterogeneous sensing nodes, data processing logic of edge computing gateways, and model update mechanism of cloud analysis platforms.

[0041] First, at the heterogeneous sensing node level, in addition to integrating a profile moisture sensor based on the frequency domain reflection principle and a surface moisture sensor based on the thermal pulse principle, this embodiment further adds a soil temperature sensor and a soil conductivity sensor. The soil temperature sensor uses a digital thermistor element, mounted on the same probe rod as the profile moisture sensor, and simultaneously measures soil temperature at depths of 5 cm, 15 cm, 30 cm, and 50 cm. The soil conductivity sensor, employing a four-electrode method, is also integrated at the bottom of the probe rod and is used to measure the salinity content of deep soil. During each sampling cycle, the microcontroller unit simultaneously collects three types of data: moisture, temperature, and conductivity, and uses a pre-stored cross-correction coefficient matrix to correct the moisture measurements for temperature and salinity. This temperature and salinity correction process is based on the following empirical formula:

[0042]

[0043] in, This is the corrected soil volumetric moisture content. These are the original measured values. To measure soil temperature, This is the standard reference temperature (usually 25 degrees Celsius). To measure soil electrical conductivity, The standard reference conductivity is (usually taken as 0.5 dS / m). and The sensitivity coefficients for temperature and electrical conductivity were obtained through laboratory calibration. By introducing a temperature and electrical conductivity compensation mechanism, the accuracy and stability of moisture measurement under different soil physicochemical conditions were significantly improved.

[0044] Secondly, in the data cleaning unit of the edge computing gateway, this embodiment introduces a dynamic anomaly detection algorithm based on a sliding time window. Traditional fixed standard deviation threshold methods are prone to misclassifying normal data as outliers when faced with rapid changes in soil moisture (such as after rainfall). Therefore, the data cleaning unit maintains a sliding time window with a length of 5 sampling periods, performing local trend fitting on the historical data sequence of each node at each depth layer. If the absolute value of the residual between the current data point and the fitted trend line exceeds twice the standard deviation of the residual within that window, it is determined to be an anomaly. This method can adaptively adjust the anomaly judgment threshold, effectively distinguishing between real rapid changes and sensor malfunctions, thus improving the intelligence level of data cleaning.

[0045] Furthermore, in the spatiotemporal co-Kriging interpolation engine of the cloud-based analytics platform, this embodiment adds an online learning and adaptive update mechanism for model parameters. Initial range parameters... Although learned through training on historical data, the spatiotemporal correlation of soil moisture varies significantly under different seasons, crop rotations, or extreme weather events. Therefore, after each interpolation calculation, the engine performs residual analysis on the covariance model using newly received profile validation point data and fine-tunes the range parameters using recursive least squares. The adjustment magnitude is constrained by a smoothing factor to prevent drastic parameter fluctuations. This online learning and adaptive update mechanism ensures that the interpolation model can continuously adapt to the dynamic evolution of the farmland environment, maintaining long-term modeling accuracy.

[0046] Finally, in the irrigation decision interface module, this embodiment introduces a multi-objective optimization strategy. In addition to meeting crop water requirement thresholds, the decision model comprehensively considers water resource quota constraints, energy consumption costs (such as pump electricity costs), and soil salinization risks. By constructing a weighted objective function, the system can solve for the optimal irrigation scheme under multiple constraints, rather than a single "whether to irrigate" decision. For example, in water-scarce areas, the system may suggest extending the irrigation interval but increasing the amount of water per irrigation to balance crop demand and resource conservation. This multi-objective optimization strategy makes irrigation decisions more flexible and economical, applicable to diverse agricultural production management objectives.

[0047] Through the above improvements, this embodiment significantly enhances its adaptability and decision-making intelligence under complex soil environments, dynamic climate conditions, and diverse management objectives while maintaining the core system architecture, further expanding the application boundaries and practical value of the present invention.

Claims

1. A smart soil moisture analysis system based on an Internet of Things (IoT) terminal, characterized in that, include: An Internet of Things (IoT) terminal network deployed in the target farmland area, the IoT terminal network consisting of multiple heterogeneous sensor nodes, each heterogeneous sensor node integrating at least two different types of soil moisture sensors, microcontroller units, low-power wide-area network communication modules, and energy harvesting and management modules; An edge computing gateway deployed at the edge of farmland is used to aggregate and preprocess data from the Internet of Things terminal network. The edge computing gateway includes a multi-protocol access unit, a data cleaning unit, and a preliminary spatial interpolation unit. A cloud-based analytics platform is used to perform deep data fusion and three-dimensional soil moisture modeling. The cloud-based analytics platform includes a data receiving and storage service, a spatiotemporal co-working kriging interpolation engine, and a three-dimensional soil moisture model construction module. The operation process of the spatiotemporal co-Kriging interpolation engine is as follows: First, the two-dimensional surface soil moisture distribution raster data of the current moment and the previous four consecutive moments are extracted from the data receiving and storage service and used as spatially continuous trend surface data. Extract the profile depth data collected by all the heterogeneous sensing nodes within the same time period, and use it as spatially discrete verification point data with vertical dimension; A composite covariance function model integrating spatial autocorrelation, temporal autocorrelation, and spatial depth correlation is constructed. This composite covariance function model coordinates the surface trend surface data and profile point data in the three-dimensional domain of spatiotemporal depth. Perform co-kriging interpolation to find the optimal unbiased estimate of soil volumetric moisture content at any horizontal position and any depth for nodes not deployed, and finally output a continuously distributed soil moisture data model with three-dimensional spatial coordinates. The multi-protocol access unit of the edge computing gateway is responsible for receiving data packets from all the heterogeneous sensing nodes and parsing out the location coordinates of each node and its corresponding multi-depth soil moisture content dataset. The cloud-based analytics platform continuously receives two-dimensional surface soil moisture distribution raster maps and raw multi-depth profile data from all nodes from the edge computing gateway, and stores them in chronological order.

2. The intelligent soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The at least two different types of soil moisture sensors specifically include a profile moisture sensor based on the frequency domain reflection principle and a surface moisture sensor based on the thermal pulse principle. The profile moisture sensor adopts a multi-probe array structure, with its probes inserted vertically into the soil at non-equidistant intervals to simultaneously measure the soil volumetric moisture content at depths of 5 cm, 15 cm, 30 cm, and 50 cm below the surface. The surface moisture sensor is installed close to the ground surface and is used to measure the soil moisture status within a depth range of 0 to 2 cm from the ground surface. The microcontroller unit periodically and synchronously acquires raw voltage signals from the profile moisture sensor and the surface moisture sensor at a preset sampling frequency, and converts the raw voltage signals into standardized soil volumetric moisture content values ​​through a built-in conversion algorithm pre-calibrated for each sensor model.

3. The intelligent soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The energy harvesting and management module includes solar photovoltaic panels, supercapacitor banks, and power management integrated circuits. The power management integrated circuit monitors the output power of the solar photovoltaic panel and the remaining power of the supercapacitor bank in real time. When the charge of the supercapacitor bank is detected to be lower than the first preset threshold, the power management integrated circuit switches the heterogeneous sensing node to an ultra-low power sleep mode, maintaining only the basic clock and wake-up timer, and stopping all sensor data acquisition and wireless communication. When the output power of the solar photovoltaic panel continuously exceeds the second preset threshold and the charge of the supercapacitor bank recovers to above the third preset threshold, the power management integrated circuit controls the heterogeneous sensing node to exit the sleep mode and resume normal operation.

4. The intelligent soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The low-power wide-area network communication module adopts a communication protocol based on spread spectrum technology. After each wake-up, the microcontroller unit encapsulates the multi-depth soil moisture content data collected and converted in the current cycle, along with the node's own GPS positioning coordinates and collection timestamp, into a data packet. The low-power wide-area network communication module transmits the data packets to the edge computing gateway with a transmission power of less than 20 milliwatts; the data packet transmission adopts a random backoff and acknowledgment retransmission mechanism. The data cleaning unit performs outlier detection and removal on the received dataset. The specific process is as follows: for the water content data of the same depth layer, calculate the average value and standard deviation of all node data. Individual data points whose values ​​deviate from the mean by more than 2 standard deviations are marked as outliers; For data points marked as outliers, replace them with the linear interpolation result of the two adjacent valid historical measurements of the data point. The preliminary spatial interpolation unit processes the cleaned surface soil moisture content data at the same sampling time and uses an inverse distance weighted interpolation algorithm to generate a two-dimensional surface soil moisture distribution raster map with a resolution of 1 meter × 1 meter covering the entire farmland area.

5. The intelligent soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The three-dimensional soil moisture model construction module receives the soil moisture data model output by the spatiotemporal co-Kriging interpolation engine and converts it into a renderable and analyzable three-dimensional data structure. The soil moisture data module can dynamically generate a high-precision two-dimensional thematic map of soil moisture distribution based on any horizontal slice depth or vertical profile position specified by the user. The soil moisture data module integrates a water transport simulation algorithm, which can simulate the redistribution of soil moisture in the next 6, 12 and 24 hours based on a three-dimensional soil moisture data model, and predict the trend of effective water content changes at different depths in the crop root zone.

6. The intelligent soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The cloud-based analysis platform is also connected to an irrigation decision interface module; The irrigation decision interface module calls the built-in irrigation decision model based on the crop root zone effective water content prediction trend provided by the three-dimensional soil moisture model construction module, combined with the pre-input crop type, growth stage and soil type parameters. The irrigation decision model automatically generates a set of control instructions, including suggested irrigation time, irrigation duration, and irrigation amount, by comparing the predicted effective water content with the crop's water requirement threshold at that growth stage. This set of instructions is then distributed to the farmland through an application programming interface (API) to form an intelligent irrigation execution system.

7. The intelligent soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The composite covariance function model is in the following form: ; in, Indicates spatial lag distance Time lag interval and depth lag difference The covariance value under the following conditions; For prior variance; , , These are the range parameters in the spatial, temporal, and depth directions, respectively.

8. A smart soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The heterogeneous sensing node also integrates a soil temperature sensor and a soil conductivity sensor; the microcontroller unit simultaneously collects three types of data—moisture, temperature, and conductivity—in each sampling cycle, and uses a pre-stored cross-correction coefficient matrix to correct the moisture measurement value for temperature and salinity. The temperature and salinity correction process is based on the following empirical formula: ； in, This is the corrected soil volumetric moisture content. These are the original measured values. To measure soil temperature, For standard reference temperature, To measure soil electrical conductivity, For standard reference conductivity, and The temperature and conductivity sensitivity coefficients were obtained through laboratory calibration.

9. A smart soil moisture analysis system based on an Internet of Things terminal according to claim 1, characterized in that, The data cleaning unit of the edge computing gateway adopts a dynamic anomaly detection algorithm based on a sliding time window. The dynamic anomaly detection algorithm maintains a sliding time window with a length of 5 sampling periods and performs local trend fitting on the historical data sequence of each node at each depth layer; if the absolute value of the residual between the current data point and the fitted trend line exceeds twice the standard deviation of the residual within the sliding time window, it is determined to be an anomaly. The spatiotemporal co-Kriging interpolation engine also includes an online learning and adaptive update mechanism for model parameters. After each interpolation calculation, this mechanism uses newly received profile verification point data to perform residual analysis on the composite covariance function model and employs recursive least squares to adjust the range parameters. , , Fine-tuning is performed, with the adjustment range constrained by the smoothing factor.

Images