Cultivated land quality monitoring method

Abstract

The invention relates to the technical field of agricultural environment monitoring, and discloses a cultivated land quality monitoring method. The method comprises the following steps: acquiring soil multi-parameter data, and carrying out sensor network deployment and anti-interference processing; performing exception filtering, standardization and missing value interpolation on the data; cloud storage data are generated through sliding window segmentation and encryption compression; generating a monitoring report based on the multi-dimensional analysis and the dynamic chart; adopting a plurality of intelligent algorithms to identify abnormal modes and constructing an early warning strategy library; and finally, real-time monitoring and early warning grade division are realized. According to the invention, full-process automatic monitoring and intelligent early warning of the cultivated land quality are realized, and the monitoring precision and the decision-making efficiency are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural environmental monitoring technology, and in particular to a method for monitoring arable land quality. Background Technology

[0002] In the field of agricultural environmental monitoring technology, existing solutions related to farmland quality monitoring typically employ fixed sensor deployment and periodic data acquisition, which suffers from limitations such as insufficient anti-interference capabilities of sensor nodes, low efficiency in multi-source data synchronous processing, and delayed dynamic early warning response. Existing methods often rely on single-parameter threshold determination or static data analysis models, which are prone to data acquisition bias accumulation, high false alarm rates for abnormal states, and insufficient timeliness of trend prediction in complex farmland real-time monitoring scenarios. These shortcomings make it difficult to meet the accuracy requirements of dynamic monitoring and early warning of farmland quality. Regarding the joint optimization of sensor network configuration, node calibration, and multi-source data processing links, existing technologies generally lack a collaborative mechanism for suppressing communication interference and extracting data correlation features. This makes it difficult to form a complete process of sensor deployment—dynamic calibration—data acquisition—trend analysis—early warning triggering under real-time monitoring and accuracy constraints, resulting in limited ability to analyze the spatial heterogeneity of farmland quality parameters and capture temporal evolution patterns. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method for monitoring arable land quality, comprising: The system acquires soil temperature, humidity, pH, and nutrient parameters. Using farmland geographic coordinates and soil type information as input, it performs multi-parameter sensor network deployment, adjusts transmit power, receive sensitivity, and channel selection strategies to perform anti-interference parameter calibration, including clock synchronization correction, and generates preprocessed sensor data. Based on preprocessed sensor data, reasonable threshold ranges for physical and chemical parameters are set for outlier filtering, the time zone and format of the unified timestamp are standardized to Coordinated Universal Time (UTC) for format conversion, and missing value imputation is performed using linear interpolation, spline interpolation, and machine learning-based predictive imputation to generate a complete sensor dataset. Based on a complete sensor dataset, the sliding window is segmented according to the preset sliding window size and step size parameters, an autoregressive moving average model is used for trend fitting analysis, and a symmetric encryption algorithm combined with a key management mechanism is used for encryption and compression processing to generate a cloud storage data structure. Based on the cloud storage data structure, multi-dimensional data parsing is performed according to the metadata fields in the cloud storage data structure to generate line charts, heat maps, radar charts and trend curves for dynamic chart rendering, and access control lists and role-based permission management mechanisms are used for permission verification to generate dynamic monitoring reports. The system acquires dynamic monitoring reports, uses isolated forest, support vector machine and cluster analysis methods for abnormal pattern recognition, employs multi-level threshold strategy for threshold comparison analysis and rule engine architecture for response rule matching, and generates an intelligent early warning strategy library. Based on the intelligent early warning strategy library, real-time data stream monitoring and early warning level classification are performed to generate sensor network configuration update parameters.

[0004] Furthermore, the outlier filtering process includes: Outlier filtering is performed by setting reasonable threshold ranges for physical and chemical parameters and comparing each data point collected by each sensor one by one. Based on the calculation of the mean and standard deviation of the sliding window, local abrupt changes and abnormal fluctuations are identified, and combined with historical data trends, it is determined whether the data points are abnormal.

[0005] Furthermore, the process of standardizing the time zone and format of unified timestamps to the Coordinated Universal Time (UTC) standard includes: Standardized format conversion uses a timestamp correction algorithm to adjust non-standard time formats and unify the time zone and format of timestamps to the Coordinated Universal Time standard. Data format differences between different sensor types are uniformly processed by adopting data type conversion, unit conversion, and coding standards.

[0006] Furthermore, the process of using linear interpolation and spline interpolation includes: Missing value imputation is handled using a variety of imputation techniques, including linear interpolation, spline interpolation, and machine learning-based predictive imputation. For short-term missing data, linear or spline interpolation methods are preferred. For long-term missing data or complex missing data patterns, time series prediction models trained on historical data are used for filling.

[0007] Furthermore, the process of segmenting the sliding window based on preset sliding window size and step parameters includes: The sliding window segmentation process divides the complete sensor dataset into time series segments based on preset sliding window size and step size parameters, forming a series of overlapping or non-overlapping time window data blocks. The size and step size of the sliding window are dynamically adjusted according to the temporal characteristics of soil parameter changes and monitoring requirements.

[0008] Furthermore, the process of employing an autoregressive moving average model includes: Trend fitting analysis includes performing feature extraction operations on multi-parameter data within each time window, and calculating statistical and temporal features; The variation trends of soil quality parameters were modeled using polynomial fitting, exponential smoothing, and autoregressive moving average models. Combining soil environmental factors and historical trends, the variation curves of each parameter were fitted.

[0009] Furthermore, the process of conducting trend fitting analysis includes: The encrypted compression process uses a symmetric encryption algorithm combined with a key management mechanism, dynamically adjusting the encryption strength and key update frequency based on the data sensitivity level and access permission policy. Lossless compression algorithms are used to efficiently compress structured time series trend data, and differential coding, entropy coding and dictionary coding are selected to optimize the compression ratio.

[0010] Furthermore, the expression for generating preprocessed sensor data includes: Calculate the optimal spacing between sensor nodes: ; in, This represents the entropy value of the deployment area; A soil type classification index; Number of soil type classifications; For the first The probability percentage of soil types in the deployment area; For the first Identifiers for soil types; An improved Shannon entropy index modulator; Cross-correlation analysis is used to optimize communication frequency band selection and calculate the frequency band interference coefficient: ; in, This represents the cross-correlation value across frequency bands. and These represent the signal waveforms of the main frequency band and the interference frequency band, respectively. For delay parameters; It is a time variable; Construct the objective function: ; in, This represents the transmit power of the sensor node; This refers to the receiving sensitivity of the sensor node; Maximum permissible power; These are the weighting coefficients; This represents the cross-correlation value across frequency bands. Interference threshold; Correcting clock synchronization errors: ; in, This represents the average clock deviation. The total number of nodes; For sensor node indexing; For the first The local timestamp of each node; As a reference time base; Define data integrity metrics: ; in, For data integrity rate; The number of valid data packets; This represents the total number of data packets; This is the time deviation attenuation coefficient; This represents the average clock offset. Execution timestamp alignment: ; in, The optimal alignment reference time is calculated; The baseline time variable to be optimized; Indexing timestamp data; For the first One timestamp to be aligned; This represents the number of nodes.

[0011] Furthermore, the expressions for generating cloud storage data structures include: Define window stability metrics: ; in, An improved Chebyshev norm for window data; A multi-parameter data matrix within a time window; This is the row index of the data matrix; For column indices of the data matrix; The dimension of the data matrix; The first in the data matrix row element value, The first in the data matrix The element values ​​of the column; For smoothing coefficients; It is a fixed label that indicates a specific parameter type or sensor type; Calculate window overlap ratio: ; in, The overlap rate; and These are the overlap duration and the window duration, respectively. To improve the Chebyshev norm; It is a regular Chebyshev norm; Construct the prediction equation: ; in, These are predicted values ​​for soil parameters; For time indexing; The order of the autoregressive model; and These are the coefficients for autoregression and moving average; This is the lag order index for the autoregressive term; The order of the moving average model; This is the lag order index for the moving average term; It is a white noise sequence; Calculate the trend confidence level: ; in, Confidence level; Index the data points; This represents the total number of data points within the current time window. For the first The fitting residuals of each data point; For the first The actual observed values ​​of each data point; Define the compression factor: ; in, To compress the Reynolds number; For data density; For data flow rate; This refers to the data block size. The coefficient of compressibility viscosity; For trend confidence.

[0012] Furthermore, the process of generating data on quality change trends includes: Feature extraction is performed on multi-parameter data within each time window, including calculating statistical features such as mean, median, variance, skewness, and kurtosis to reflect the central tendency and distribution characteristics of soil parameters; extracting temporal features such as maximum, minimum, and rate of change to describe the dynamic changes of parameters; calculating cumulative content and variation amplitude for nutrient parameters to assess the temporal evolution of soil nutrients; and combining sensor data quality labels to weight abnormal or missing data to reduce the impact of outliers on feature calculation.

[0013] The key innovations of this invention include: (1) The sensor network configuration is dynamically optimized by using improved Shannon entropy and cross-correlation analysis techniques, involving multi-source data processing of soil temperature, humidity, pH value and nutrient parameters.

[0014] (2) An anti-interference parameter calibration mechanism is adopted to collect data from the calibration sensor nodes in real time to improve the accuracy and stability of the data.

[0015] (3) Apply sliding window segmentation processing technology to perform trend fitting analysis on the complete sensor dataset, generate quality change trend data and encrypt and compress it to ensure data security and timeliness.

[0016] The following are its main beneficial effects: (1) By dynamically optimizing the configuration of the sensor network, the collaborative work of multiple parameters sensors was realized, which enhanced the data acquisition capability in complex farmland environments and reduced the accumulation of data acquisition deviations.

[0017] (2) The application of the anti-interference parameter calibration mechanism enables the calibration sensor node to accurately collect real-time data even when the anti-interference capability is insufficient, thereby reducing the false judgment rate of abnormal state.

[0018] (3) The use of sliding window segmentation technology makes the generation of quality change trend data more timely and accurate, ensuring the ability to capture the temporal evolution pattern of cultivated land quality monitoring, and improving the security of data storage through encryption compression. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a method for monitoring farmland quality provided in an embodiment of this application. Detailed Implementation

[0020] Example 1: Refer to Figure 1 This is a flowchart illustrating a method for monitoring arable land quality provided in an embodiment of the present invention. The process may include at least steps S100-S600: S100: Acquire soil temperature, humidity, pH value, and nutrient parameters; use farmland geographic coordinates and soil type information as input to deploy a multi-parameter sensor network; adjust transmission power, receiver sensitivity, and channel selection strategies to perform anti-interference parameter calibration, including clock synchronization correction, and generate preprocessed sensor data. S200: Based on preprocessed sensor data, set reasonable threshold ranges for physical and chemical parameters to filter outliers, standardize the time zone and format of the unified timestamp to Coordinated Universal Time (UTC) standard for format conversion, and use linear interpolation, spline interpolation, and machine learning-based predictive interpolation to impute missing values, thereby generating a complete sensor dataset. S300: Based on a complete sensor dataset, the system performs sliding window segmentation according to preset sliding window size and step size parameters, uses an autoregressive moving average model for trend fitting analysis, and employs a symmetric encryption algorithm combined with a key management mechanism for encryption and compression processing to generate a cloud storage data structure. S400, based on cloud storage data structure, performs multi-dimensional data parsing based on metadata fields in cloud storage data structure, generates line charts, heat maps, radar charts and trend curves for dynamic chart rendering, and uses access control lists and role-based permission management mechanisms for permission verification, and generates dynamic monitoring reports. S500 acquires dynamic monitoring reports, uses isolated forest, support vector machine and cluster analysis methods for abnormal pattern recognition, uses multi-level threshold strategy for threshold comparison analysis and uses rule engine architecture for response rule matching processing, and generates an intelligent early warning strategy library; The S600, based on an intelligent early warning strategy library, performs real-time data stream monitoring and early warning level classification, and generates sensor network configuration update parameters.

[0021] Step S100 includes at least steps S110-S130: S110. Obtain soil temperature, humidity, pH value, and nutrient parameters. Deploy a multi-parameter sensor network using farmland geographic coordinates and soil type information as input. Adjust the transmit power, receive sensitivity, and channel selection strategy to perform anti-interference parameter calibration, including clock synchronization correction, and obtain the sensor network configuration. Specifically, the sensor network deployment first takes the farmland environment as its foundation, selecting suitable soil temperature and humidity sensors, pH sensors, and nutrient sensors as components of a multi-parameter sensor system. This multi-parameter sensor system can perceive key physical and chemical indicators of the soil in real time. This step uses pre-set farmland geographical coordinates and soil type information as input, combined with actual on-site deployment requirements, to determine the number, location, and spacing of sensor nodes, forming a preliminary sensor deployment plan. Specifically, the installation of sensor nodes includes physical fixing, power supply system setup, and implementation of environmental protection measures to ensure stable operation of the sensors in the farmland environment. Further, based on the technical specifications of the multi-parameter sensors, the initial sampling frequency, data acquisition cycle, and sensor identification information are configured to complete the basic parameter settings of the sensor nodes. The sensor network configuration includes the physical location coordinates of the sensor nodes, sensor type and parameter settings, sampling time interval, and network communication interface parameters. Throughout the deployment process, the system automatically records the status information and installation time of each node, generating a sensor network configuration data structure as the input basis for subsequent steps. This sensor network configuration is uploaded to the processing center via a wireless or wired communication link for subsequent anti-interference parameter calibration. The sensor network configuration is used as the output field name for this step, and will be called by "Sensor Network Configuration" in S120 to realize the calibration and optimization of sensor nodes.

[0022] S120. Extract communication parameters from the sensor network configuration, perform anti-interference parameter calibration, and generate calibration sensor nodes; Specifically, the communication parameters in the sensor network configuration are used as input to comprehensively analyze the wireless signal transmission quality, signal strength, and environmental interference factors of the sensor nodes, and to perform an anti-interference parameter calibration process. This process first collects electromagnetic interference background data of the sensor node's environment, and then identifies potential interference sources and patterns based on the sensor node's communication frequency band and signal modulation method. Further, based on the interference patterns, the system dynamically adjusts the sensor node's transmit power, receive sensitivity, and channel selection strategy to optimize the stability of the communication link and the reliability of data transmission. Specifically, the system monitors the communication bit error rate and packet loss rate through a feedback mechanism, adjusting anti-interference parameters in real time to form a closed-loop regulation. Anti-interference parameter calibration also includes correcting the internal clock synchronization mechanism of the sensor nodes to ensure the accuracy and consistency of data acquisition timestamps. During calibration, abnormal nodes and communication anomalies are automatically recorded, triggering the activation of backup communication paths to ensure data continuity. After completing the anti-interference parameter calibration, the sensor node's configuration file is updated to generate a calibrated sensor node data structure, including optimized communication parameters, node status, and calibration timestamps. The calibration sensor node is used as the output field name for this step and will be subsequently called by "calibration sensor node" in S130 for high-quality data input for real-time data acquisition.

[0023] S130. Real-time data acquisition of the calibration sensor nodes to generate preprocessed sensor data; Specifically, based on the calibrated sensor nodes, a real-time data acquisition process for the multi-parameter sensor network is initiated. The system periodically acquires parameters such as soil temperature, humidity, pH, and nutrients according to the calibrated sampling frequency and time synchronization mechanism. During data acquisition, sensor nodes upload raw sensor readings to the local aggregation node or directly to the processing center via the configured communication link, ensuring the real-time performance and integrity of data transmission. Further, the acquired raw data undergoes preliminary format conversion, including standardizing data units, aligning timestamps, and parsing data packets, forming a structured sensor data stream. The acquisition system incorporates an anomaly detection mechanism to monitor the reasonable range of sensor readings and communication status in real time. Abnormal data is marked and recorded for subsequent cleaning. The data acquisition process supports multi-threaded concurrent operation, ensuring efficient processing of large-scale sensor network data. The resulting structured data set is the preprocessed sensor data, containing time-series multi-parameter measurements and corresponding metadata. This preprocessed sensor data serves as the output field name for this step and is subsequently called by "Preprocessed Sensor Data" in S210 for anomaly filtering and further cleaning of the sensor data.

[0024] In another embodiment, using the farmland geographic coordinates and soil type information as input, the system optimizes the deployment density of sensor nodes based on an improved Shannon entropy. Formula ① calculates the optimal spacing between sensor nodes: ; in: The entropy value representing the deployment area is calculated from soil type distribution data in the sensor network configuration. A soil type classification index; Number of soil type classifications; For the first The probability percentage of soil types in the deployment area; For the first Identifiers for soil types; This is an improved Shannon entropy index adjustment factor.

[0025] Formula ① outputs the entropy value to dynamically adjust the node spacing, ensuring that dense nodes are deployed in highly heterogeneous areas.

[0026] Further cross-correlation analysis is used to optimize the selection of communication frequency bands, and formula ② is used to calculate the frequency band interference coefficient: ; in: The cross-correlation value for the frequency band is calculated from the communication parameters extracted from the sensor network configuration; and These represent the signal waveforms of the main frequency band and the interference frequency band, respectively. For delay parameters; It is a time variable.

[0027] Formula ② outputs the frequency band with the lowest cross-correlation value, forming the basis for anti-interference parameter calibration. The output field in this step is named "Sensor Network Configuration," and is called by S120's "Sensor Network Configuration."

[0028] Entropy value inherited from S110 cross-correlation value The S120 employs quadratic programming to optimize transmit power and receive sensitivity. Formula ③ constructs the objective function: ; in: This represents the transmit power of the sensor node; This refers to the receiving sensitivity of the sensor node; Maximum permissible power; These are the weighting coefficients; The cross-correlation value of the frequency band (calculated by formula ②); This is the interference threshold.

[0029] Formula ③ is obtained by solving for the optimal solution. and This minimizes the bit error rate of the communication link.

[0030] Formula ④ Corrects clock synchronization error: ; in: The average clock offset is calculated from the clock data of the calibrated sensor nodes; The total number of nodes; For sensor node indexing; For the first The local timestamp of each node; Used as a reference time base.

[0031] Formula ④ outputs parameters used to dynamically adjust the clock synchronization protocol and generate a calibration sensor node. The output field for this step is named "Calibrate Sensor Node" and is called by S130's "Calibrate Sensor Node".

[0032] Optimized parameters based on S120 , and clock deviation The S130 uses a sliding window mechanism to collect data. Formula ⑤ defines the data integrity index: ; in: To assess data integrity, evaluate the validity and timeliness of data within a collection period; The number of valid data packets; This represents the total number of data packets; This is the time deviation attenuation coefficient; This represents the average clock offset.

[0033] Formula ⑤ is used to evaluate the quality of data collection. The data retransmission mechanism is triggered in time.

[0034] Formula ⑥ Execution timestamp alignment: ; in: The optimal alignment reference time is calculated from the timestamp sequence of the calibrated sensor nodes. calculate; The baseline time variable to be optimized; Indexing timestamp data; For the first One timestamp to be aligned; This represents the number of nodes.

[0035] Formula 6 outputs preprocessed sensor data used to generate time synchronization.

[0036] This section summarizes the technical effects: By using entropy optimization and cross-correlation analysis, the sensor network achieves interference-resistant deployment; combined with quadratic planning and clock correction, the integrity of data acquisition is ensured, providing highly reliable input for subsequent processing.

[0037] Step S200 includes at least steps S210-S230: S210. Acquire preprocessed sensor data, set reasonable physical and chemical parameter threshold ranges to perform outlier filtering, and obtain cleaned sensor data. Specifically, the preprocessed sensor data serves as the input source, comprising a structured data stream of soil temperature, humidity, pH, and nutrient parameters collected by a multi-parameter sensor network. The system first performs outlier filtering on this data stream. This process compares each data point collected by each sensor against reasonable threshold ranges for physical and chemical parameters, eliminating outliers that significantly exceed the actual environmental range. Furthermore, statistical methods, such as mean and standard deviation calculations based on a sliding window, are used to identify local abrupt changes and abnormal fluctuations, and combined with historical data trends, to determine whether data points are abnormal. Outlier detection also incorporates sensor node status information, such as communication interruptions, data packet loss, and sensor fault identifiers, to comprehensively determine data validity. For data determined to be abnormal, the system uses a tagging mechanism to annotate it, and records the abnormal events and their timestamps in an anomaly log database for subsequent analysis. This outlier filtering process supports multi-threaded parallel processing, ensuring real-time performance in large-scale sensor network environments. The filtered data will either have outliers removed or be temporarily replaced by interpolation methods, forming a cleaned sensor data set. The cleaned sensor data structure includes multi-parameter time series data and corresponding anomaly identification fields, ensuring data integrity and accuracy. Ultimately, the cleaned sensor data serves as the output field name for this step, subsequently called by "Cleaned Sensor Data" in S220 for standardized format conversion and timestamp sequence extraction, supporting the generation of time series datasets.

[0038] S220. Extract timestamp sequences from the cleaned sensor data, standardize the time zone and format of the timestamps to Coordinated Universal Time (UTC) standard, and generate a time series dataset. Specifically, using the cleaned sensor data as input, the system first extracts a complete and continuous timestamp sequence. Based on the sensor sampling frequency and time synchronization mechanism, it identifies missing, duplicate, and incorrect timestamps. This step uses a timestamp correction algorithm to adjust non-standard time formats, unifying the timestamp time zone and format to Coordinated Universal Time (UTC) to ensure time alignment across nodes. Further, based on the timestamp sequence, the multi-parameter sensor data is sorted along a time axis and mapped to a unified time series data structure. This structure uses a standardized data format description, including timestamp fields, multi-parameter measurement value fields, and data quality identifier fields, facilitating subsequent algorithm processing. During the standardized format conversion process, the system uniformly handles data format differences between different sensor types, employing data type conversion, unit conversion, and encoding standards to ensure data compatibility and consistency. For records with missing or abnormal timestamps, the system uses interpolation or adjacent data filling strategies within time windows to reduce time series gaps. This process also generates metadata for the time series dataset, such as sampling interval statistics, data integrity rate, and data quality indicators, for monitoring and optimization. The time series dataset is used as the output field name for this step. It is subsequently called by "Time Series Dataset" in S230 for missing value imputation and generation of complete datasets, while also supporting cross-step data analysis and trend prediction.

[0039] S230. Use linear interpolation, spline interpolation and machine learning-based predictive imputation to impute missing values ​​in the time series dataset to generate a complete sensor dataset. Specifically, the system uses the time-series dataset as input and performs imputation processing on missing values ​​and discontinuous data to generate a complete sensor dataset. This imputation process first identifies missing intervals in the time series and selects an appropriate imputation algorithm based on the length of the missing data and upstream / downstream data trends. Specifically, multiple imputation techniques are employed, including linear interpolation, spline interpolation, and machine learning-based predictive imputation, dynamically selected for different missing patterns. For short-term missing data, the system prioritizes linear or spline interpolation methods to maintain data continuity; for long-term missing data or complex missing patterns, a time-series prediction model trained on historical data is used for imputation, incorporating soil parameter variation patterns and environmental influencing factors to improve the accuracy of the imputed data. During the imputation process, the system evaluates the confidence level of the imputation results in real time, using residual analysis and cross-validation to ensure that the imputed data conforms to the actual trend. Furthermore, the imputed data is fused with the original data, updating the data quality identifier in the time-series data structure to distinguish between the original observation data and the imputed data. This complete sensor dataset contains continuous and missing multi-parameter time-series data, meeting the input requirements for subsequent sliding window segmentation and trend fitting analysis. The complete sensor dataset is used as the output field name for this step. It is subsequently called by "Complete Sensor Dataset" in S310 for the generation of time window data blocks and in-depth analysis. It also supports data storage and visualization processing across main steps.

[0040] Step S300 includes at least steps S310-S330: S310. Obtain the complete sensor dataset, and perform sliding window segmentation processing according to the preset sliding window size and step size parameters to obtain time window data blocks; Specifically, the complete sensor dataset serves as the input source, containing continuous multi-parameter time-series data after imputation of missing values. The system first performs time-series segmentation on the complete sensor dataset based on preset sliding window size and step size parameters. This sliding window segmentation process moves the window position gradually along the time axis in fixed-length time intervals according to a predetermined time step, forming a series of overlapping or non-overlapping time window data blocks. Specifically, the system calculates the start and end times of each window based on the sampling frequency and timestamp information of the time-series dataset, ensuring that the time window covers continuous and complete data segments. The sliding window size and step size parameters can be dynamically adjusted according to the temporal characteristics of soil parameter changes and monitoring needs to adapt to different analytical granularity and real-time requirements. Further, for each time window, the system extracts multi-parameter measurement values ​​within the corresponding time period from the complete sensor dataset, forming a time window data block containing multi-dimensional data such as soil temperature and humidity, pH value, and nutrients. During this extraction process, the system checks the completeness and quality indicators of the data within the time window. If data is found to be missing or the proportion of abnormalities exceeds a threshold, the time window data block is marked for reference by subsequent processing modules. The time window data block also includes corresponding timestamp ranges and data quality descriptions, supporting subsequent feature extraction and trend analysis. The sliding window segmentation process employs multi-threaded parallel execution to meet the real-time processing requirements of large-scale sensor network data. The time window data block serves as the output field name for this step and is subsequently called by S320's "Time Window Data Block" for feature vector extraction and trend fitting analysis, while also supporting data storage and visualization across main steps.

[0041] S320. Extract feature vectors from the time window data block, use an autoregressive moving average model to perform trend fitting analysis, and generate quality change trend data. Specifically, using the time window data blocks as input, the system first performs feature extraction on the multi-parameter data within each time window. This feature extraction includes calculating statistical features such as mean, median, variance, skewness, and kurtosis to reflect the central tendency and distribution characteristics of soil parameters; further, it extracts time-domain features such as maximum, minimum, and rate of change to describe the dynamic changes of the parameters. For nutrient parameters, the system also calculates the cumulative content and magnitude of change to assess the temporal evolution of soil nutrients. During feature extraction, the system combines the quality identifiers of sensor data to weight abnormal or missing data, reducing the impact of outliers on feature calculation. Subsequently, based on the feature vectors, the system models the changing trends of soil quality parameters using a trend fitting analysis algorithm. Specifically, it employs time series analysis methods such as multinomial fitting, exponential smoothing, and autoregressive moving average (ARMA) models, combined with soil environmental factors and historical trends, to fit the change curves of each parameter. During trend fitting, the system dynamically adjusts the model parameters based on the fitting error and confidence interval to optimize the fitting accuracy. Furthermore, the system performs trend classification on the fitting results, identifying rising, falling, and stable change patterns to provide a basis for subsequent quality assessment and early warning. Trend fitting analysis also includes anomaly trend detection, identifying possible abrupt changes and abnormal fluctuations, and generating trend anomaly markers. Finally, the system integrates the trend fitting results and related statistical features to form structured quality change trend data, including parameter trend curves, trend classifications, and anomaly markers. This quality change trend data serves as the output field name for this step and is subsequently accessed by the "Quality Change Trend Data" function in S330 for encryption, compression, and cloud storage. It also supports the generation of visual reports and the construction of intelligent early warning models across main steps.

[0042] S330. The quality change trend data is encrypted and compressed using a symmetric encryption algorithm combined with a key management mechanism to generate a cloud storage data structure. Specifically, the system takes the quality change trend data as input and first encrypts it using a symmetric encryption algorithm combined with a key management mechanism to ensure data security during transmission and storage. During encryption, the system dynamically adjusts the encryption strength and key update frequency based on data sensitivity levels and access permission policies to prevent data leakage and unauthorized access. The encrypted data then enters the compression stage. The system uses a lossless compression algorithm to efficiently compress structured time-series trend data, reducing storage space requirements and improving transmission efficiency. During compression, the system selects a suitable compression encoding method, such as differential encoding, entropy encoding, and dictionary encoding, based on data type and distribution characteristics to optimize the compression ratio. Further, the system performs integrity verification on the encrypted and compressed data, generating a checksum for subsequent data verification and anomaly detection. Subsequently, according to the cloud platform's data storage specifications, the system organizes the encrypted and compressed data into a cloud storage data structure. This structure includes data content, metadata (such as timestamps, sensor node identifiers, and data quality indicators), access control information, and a checksum. The cloud storage data structure supports distributed storage and multi-level caching mechanisms to meet the needs of efficient management and fast access to large-scale data. The data structure design is compatible with multiple cloud service interfaces, supporting subsequent data parsing, visualization, and intelligent analysis. The cloud storage data structure serves as the output field name for this step and is subsequently called by S410's "Cloud Storage Data Structure" for multi-dimensional analysis and visualization report generation of farmland quality trends. It also supports the construction of intelligent early warning models and real-time monitoring feedback control across main steps.

[0043] In another embodiment, using the complete sensor dataset as input, S310 employs an improved Chebyshev norm to segment the time window. Equation ③ defines the window stability metric: ; in: The improved Chebyshev norm for windowed data is derived from multi-parameter time series data in the complete sensor dataset. calculate; A multi-parameter data matrix within a time window; This is the row index of the data matrix; For column indices of the data matrix; The dimension of the data matrix; The first in the data matrix The element value of the row; For smoothing coefficients; It is a fixed label that indicates a specific parameter type or sensor type.

[0044] Formula ③ is used to dynamically adjust the window size, and triggers a split when the norm exceeds the threshold.

[0045] Formula ④ calculates the window overlap rate: ; in: The overlap rate; and These are the overlap duration and the window duration, respectively. The improved Chebyshev norm calculated for formula ③; It is the regular Chebyshev norm (i.e., the maximum norm).

[0046] Formula ④ outputs a time window data block. The output field name for this step is "Time Window Data Block," which is called by S320's "Time Window Data Block."

[0047] Following the window data block from S310, S320 uses an ARMA model for trend fitting. Formula ⑤ constructs the prediction equation: ; in: These are predicted values ​​for soil parameters, calculated from the historical sequence of data blocks within a time window. For time indexing; The order of the autoregressive (AR) model; and These are the coefficients for autoregression and moving average; This is the lag order index for the autoregressive term; This represents the order of the moving average (MA) model. This is the lag order index for the moving average term; It is a white noise sequence.

[0048] Formula ⑤ outputs the fitting residuals to evaluate trend stability.

[0049] Formula 6 calculates the trend confidence level: ; in: The confidence level is an indicator used to evaluate the quality of the ARMA model fit. The closer the ratio is to 1, the better the fit and the more reliable the trend. This serves as an index for data points, used to traverse all data points within the window; This represents the total number of data points within the current time window. For the first The fitting residuals of each data point; For the first The actual observed values ​​of each data point.

[0050] Formula ⑥ outputs data used to generate quality change trend data. The output field name for this step is "Quality Change Trend Data," which is called by "Quality Change Trend Data" in S330.

[0051] Furthermore, building upon the trend data from S320, S330 employs an improved Reynolds number model to optimize the compression ratio. Formula ⑦ defines the compression factor: ; in: To compress the Reynolds number; Data density (calculated from the sampling frequency of the quality change trend data); For data flow rate; This refers to the data block size. The coefficient of compressibility viscosity; For trend confidence.

[0052] Formula ⑦ outputs the parameters of the compression algorithm to dynamically adjust the data structure for cloud storage.

[0053] This section summarizes the technical effects: By improving the Chebyshev norm and ARMA model, accurate trend fitting is achieved; combined with Reynolds number optimization compression strategy, secure data storage and efficient transmission are ensured.

[0054] Step S400 includes at least steps S410-S430: S410. Obtain the cloud storage data structure, perform multi-dimensional data parsing based on the metadata fields in the cloud storage data structure, and obtain the basic data for visualization. The system first performs multi-dimensional data parsing on the cloud storage data structure as input. Specifically, the parsing process extracts timestamp information, sensor node identifiers, data quality indicators, and access control information based on the metadata fields in the cloud storage data structure, establishing a multi-dimensional data index system. This index system supports flexible retrieval and aggregation analysis based on time, space, and parameter categories. Further, the system performs decryption and decompression operations on the encrypted and compressed data content to restore the original quality change trend data. The decryption process uses a pre-configured key management mechanism to ensure compliance of data access permissions, while decompression performs lossless restoration based on the compression algorithm type. After data restoration, the system converts the quality change trend data into a format suitable for visualization, including time series curve data, trend classification labels, and anomaly marker information. This data model supports cross-comparison and dynamic filtering of multi-dimensional data, facilitating subsequent graphic rendering and interactive operations. During the parsing process, the system, combined with data quality indicators, marks low-quality or abnormal data segments and generates a data integrity report for real-time feedback from the monitoring system. Abnormal data and access anomalies are automatically recorded in the security log database to ensure data security and audit requirements. The results of the multi-dimensional data parsing are the basic data for visualization, which serves as the output field name for this step. This data is subsequently called by the S420's "Basic Visualization Data" for extracting key quality indicators and rendering dynamic charts. It also supports the construction of intelligent early warning models and real-time monitoring and feedback control across main steps.

[0055] S420: Extract key quality indicators from visualized basic data, generate line charts, heatmaps, radar charts and trend curves for dynamic chart rendering, and generate visualized report templates; Based on the aforementioned visualization data as input, the system first extracts key quality indicators for arable land. This extraction process follows a pre-defined indicator system, including core parameters such as soil temperature and humidity, pH value, nutrient content, and their changing trends. Combined with trend classification and anomaly marker information, a set of highly representative and high-quality indicators is selected. Specifically, the system uses a weighted algorithm to comprehensively consider the temporal stability, fluctuation range, and historical trends of each indicator, generating numerical values ​​and status descriptions for the key quality indicators. Further, for these key quality indicators, the system invokes a dynamic chart rendering engine to generate various visualization chart templates based on time-series data and spatial distribution information, including line charts, heatmaps, radar charts, and trend curves. The dynamic chart rendering employs a high-performance graphics processing framework, supporting user interactive operations such as zooming, filtering, and timeline adjustment, enabling dynamic data display and real-time updates. During rendering, the system uses data quality identifiers to specially mark or prompt abnormal data points, enhancing the readability and accuracy of the report. The chart template design conforms to user permissions and access policies, supporting multi-terminal adaptation, including PC and mobile devices. The rendered visualization report template includes chart configuration parameters, data binding information, and interactive logic scripts, providing a foundation for subsequent permission verification and report generation. This visualization report template serves as the output field name for this step and is subsequently called by S430's "Visualization Report Template" for permission verification processing and dynamic monitoring report generation. It also supports intelligent early warning model construction and real-time monitoring feedback control across main steps.

[0056] S430. Use access control lists and role-based access control mechanisms to perform permission verification on the visualization report template and generate dynamic monitoring reports. Upon receiving the visualization report template as input, the system first performs a permission verification process. This process verifies the legitimacy of the current request to access the visualization report template based on the user's authentication information and access permission configuration. Specifically, the system uses Access Control Lists (ACLs) and Role-Based Access Control (RBAC) mechanisms to determine whether the user has viewing, editing, or downloading permissions, preventing unauthorized access and data leakage. During the permission verification process, the system incorporates dynamic security policies to monitor and record abnormal access behaviors in real time, triggering security alarms and restricting operation permissions. After successful verification, the system binds the visualization report template to the real-time data stream, dynamically generating a dynamic monitoring report. This report integrates dynamic charts of multi-parameter soil quality indicators, trend analysis results, and anomaly warning information, supporting customized content display and format output. During report generation, the system automatically adjusts the report layout and resolution according to user needs and device type to ensure complete information presentation and readability. Furthermore, the system adds metadata tags to the dynamic monitoring report, including generation time, data source, version information, and permission descriptions, facilitating subsequent management and tracking. The generated dynamic monitoring report is stored in the secure storage area of ​​the cloud platform, supporting historical version management and quick retrieval. This dynamic monitoring report serves as the output field name for this step and is subsequently called by the S510's "Dynamic Monitoring Report" for anomaly pattern recognition and intelligent early warning model construction. It also supports real-time monitoring feedback control and data storage maintenance across main steps.

[0057] Step 500 includes at least steps S510-S530: S510. Obtain dynamic monitoring reports, and use isolated forest, support vector machine and cluster analysis methods to perform abnormal pattern recognition processing to obtain potential abnormal data points; Using the dynamic monitoring report as input, the system first performs anomaly pattern recognition processing on the multi-parameter soil quality index data and trend analysis results in the report. Specifically, the data structure in the dynamic monitoring report is parsed into analyzable time series and spatial distribution data. Combined with the anomaly markers and trend classification information included in the report, the input dataset for the anomaly recognition model is constructed. The anomaly pattern recognition processing employs a multi-level data filtering mechanism. First, it performs preliminary screening of the fluctuation amplitude, abrupt change points, and anomalies in the time series of each parameter to identify potential anomalous data intervals. Further, the system uses machine learning-based anomaly detection algorithms, including Isolation Forest, Support Vector Machine, and clustering analysis methods, to perform in-depth analysis of the screened anomalous intervals, determining the type and severity of the anomaly patterns. Anomaly pattern recognition also incorporates environmental background information and historical anomaly event records to enhance the accuracy and robustness of the identification. During the identification process, the system dynamically adjusts the detection threshold to adapt to the characteristics of different cultivated land areas and soil types, reducing false alarms and false negatives. For identified potential abnormal data points, the system automatically generates an abnormal event description and timestamp information, and records them in the abnormal event database to support subsequent tracking and tracing. The potential abnormal data points are used as the output field names of this step, and are subsequently called by S520's "Potential Abnormal Data Points" for early warning feature extraction and threshold comparison analysis. It also supports the construction of intelligent early warning strategies and real-time monitoring feedback control across main steps.

[0058] S520. Extract early warning features from potential abnormal data points, perform threshold comparison analysis using a multi-level threshold strategy, and generate early warning triggering conditions. Based on the potential anomaly data points as input, the system first extracts early warning features. This process includes a comprehensive analysis of the spatial distribution characteristics, temporal duration characteristics, and parameter fluctuation amplitude of the anomaly data points. Specifically, the system clusters the geographical locations of the anomaly data points to identify spatial concentration areas and diffusion trends of anomaly events; simultaneously, it analyzes the duration and frequency of anomalies to assess the urgency and impact range of the anomalies. Further, combined with a soil quality threshold system, the system performs threshold comparison analysis on various parameters of the anomaly data points to determine whether they exceed preset safety ranges. The threshold comparison employs a multi-level threshold strategy, including warning thresholds, early warning thresholds, and severe thresholds, each corresponding to different risk levels. During the comparison process, the system considers the correlation and interaction effects between parameters and uses a weighted comprehensive scoring method to improve the accuracy of early warning determination. For different types of anomaly features, the system dynamically adjusts the threshold parameters to adapt to environmental changes and farmland management needs. Early warning feature extraction also includes contextual information about the anomaly events, such as sensor node status, meteorological conditions, and historical early warning records, enriching the expression of early warning triggering conditions. The system encodes the generated early warning triggering conditions in a structured format, including the anomaly event category, risk level, trigger time, and related parameter values. The warning trigger condition is used as the output field name of this step. It is subsequently called by "Warning Trigger Condition" of S530 for the generation of intelligent warning strategy library and response rule matching. It also supports real-time monitoring feedback control and data storage maintenance across main steps.

[0059] S530: The rule engine architecture is used to match response rules for the early warning trigger conditions, generating an intelligent early warning strategy library; Taking the aforementioned warning trigger conditions as input, the system first performs response rule matching processing, executing strategy matching and decision generation based on a pre-set intelligent warning strategy library. Specifically, the system compares the anomaly category, risk level, and related parameter values ​​in the warning trigger conditions with the rule set in the intelligent warning strategy library one by one, identifying warning response strategies that meet the conditions. The warning strategy library adopts a rule engine architecture, supporting condition-based trigger rules, priority sorting, and multi-rule combinations to ensure the flexibility and accuracy of response strategies. During the matching process, the system dynamically adjusts rule weights and trigger thresholds based on historical warning response effects and feedback data, improving the adaptability and intelligence level of the strategies. Further, the system generates intelligent warning strategies based on the matching results, including warning level classification, response measure suggestions, and a list of notification recipients. The generated intelligent warning strategies support multi-level linkage responses, covering sensor network configuration adjustments, on-site management measures, and remote monitoring instructions. Response rule matching also includes fault-tolerant handling for abnormal situations, such as strategy conflict detection, priority conflict resolution, and redundant strategy merging, ensuring the stability and consistency of the warning response. The system stores the generated intelligent warning strategy library in a standardized data structure, including strategy number, trigger conditions, response actions, and execution status information. The intelligent early warning strategy library is used as the output field name of this step. It is subsequently called by the "intelligent early warning strategy library" of S610 for anomaly detection and generation of hierarchical response instructions for real-time data stream monitoring. It also supports real-time monitoring feedback control and data security management across main steps.

[0060] Step S600 includes at least steps S610-S630: S610: Obtain the intelligent early warning strategy library, perform real-time data stream monitoring, and obtain anomaly detection results; The system first activates the real-time data stream monitoring module, using the intelligent early warning strategy library as input, to continuously receive real-time soil quality data from the sensor network. Specifically, the intelligent early warning strategy library contains preset early warning response rules, risk level classifications, and corresponding response measures. The system dynamically matches and analyzes the real-time data stream based on the strategies in this library. Real-time data stream monitoring accesses the latest data uploaded by sensor nodes through a data acquisition interface. Combining timestamps and sensor identification information, the system segments and buffers the data stream to ensure data continuity and temporal integrity. The system employs a high-performance stream processing engine to parse incoming data packets in real time, performing preprocessing operations including data format verification, anomaly marking, and data synchronization to form an input data stream that meets the early warning rule matching requirements. Furthermore, based on the rule set in the intelligent early warning strategy library, the system performs conditional judgments and threshold detections on various soil parameters in the real-time data stream to identify whether there are abnormal signals that meet the early warning conditions. This judgment process, combined with a multi-parameter joint analysis strategy, supports comprehensive judgment of single-parameter anomalies and multi-parameter joint anomalies, improving the accuracy and sensitivity of anomaly detection. The system records detected abnormal events in real time, including abnormal parameter values, timestamps, sensor node locations, and abnormal levels, forming an abnormal detection result data structure. The abnormal detection results are transmitted to subsequent processing modules via message queues or event-driven mechanisms to ensure the timeliness and reliability of data transmission. The abnormal detection results serve as the output field name for this step, subsequently called by S620's "Abnormal Detection Result" function for early warning level classification and generation of graded response instructions. It also supports intelligent early warning model construction and real-time monitoring feedback control across main steps.

[0061] S620. Extract response parameters from anomaly detection results, classify early warning levels, and generate graded response instructions; Based on the anomaly detection results as input, the system first extracts and structures the response parameters from the anomaly detection results. Specifically, the response parameters include the severity, impact range, duration, and specific values ​​of related soil parameters of the anomaly event. The system constructs the input feature set of the graded response judgment model by parsing the anomaly detection result data structure. Subsequently, the system performs a multi-dimensional comprehensive evaluation of the response parameters according to predefined early warning level classification standards. This evaluation combines quantitative indicators and a rule engine, classifying anomaly events into different early warning levels, such as warning, alert, and severe alert, based on the threshold range and risk level classification of the anomaly parameters. During the level classification process, the system considers the spatial distribution characteristics and temporal persistence of the anomaly event, adopts a weighted scoring mechanism, and dynamically adjusts the level classification threshold to adapt to different farmland environments and monitoring needs. Furthermore, the system generates corresponding graded response instructions based on the classified early warning levels, including response measure suggestions, a list of notification recipients, and execution priorities. The response instructions support a multi-level linkage response mechanism, covering sensor network configuration adjustments, on-site management operations, and remote monitoring instruction issuance. The system encodes tiered response commands in a standardized data format, including command number, trigger condition, execution time, and status information, facilitating subsequent tracking and management. During the command generation process, the system performs legality verification and conflict detection on the command content to avoid logical and resource conflicts between commands. The generated tiered response commands are transmitted to the feedback control module through a secure communication channel, ensuring accurate transmission and execution of the commands. The tiered response command serves as the output field name for this step and is subsequently called by the "Tiered Response Command" function of S630 for feedback control processing and sensor network configuration updates. It also supports real-time monitoring feedback control and intelligent early warning strategy optimization across main steps.

[0062] S630 performs feedback control processing on the hierarchical response commands to generate sensor network configuration update parameters.

[0063] The system receives the hierarchical response command as input and first parses and prioritizes the command content to ensure the orderly execution of feedback control. Specifically, the feedback control processing module schedules the corresponding control flow based on the execution priority and response measures in the hierarchical response command, covering the dynamic adjustment of sensor network configuration and the issuance of management commands. For sensor network configuration updates, the system extracts communication parameter adjustments, sampling frequency changes, and node activation / sleep state control involved in the command, and generates configuration update parameters based on the current operating status of the sensor nodes and the network topology. The generation process of configuration update parameters comprehensively considers network load balancing, data acquisition quality, and energy consumption optimization, employing distributed control algorithms and feedback mechanisms to dynamically adjust the working parameters of the sensor nodes. Furthermore, the system performs legality verification on the configuration update parameters, checking the correctness of the parameter range and format to prevent network failures caused by configuration anomalies. The configuration update parameters are sent to the corresponding sensor nodes through security authentication and encryption mechanisms, supporting wireless or wired communication methods to ensure the secure transmission and accurate reception of configuration commands. After receiving the configuration update parameters, the sensor nodes execute parameter updates according to their internal control logic, adjusting the communication frequency, transmission power, and data acquisition cycle to achieve dynamic optimization of network performance. The system synchronously updates the sensor network configuration database, recording configuration change history and execution status to support subsequent monitoring and auditing. Furthermore, the feedback control processing module monitors the execution effect of configuration updates, collects node feedback information, and forms a closed-loop control mechanism to ensure the effectiveness and stability of configuration adjustments. Abnormal feedback and configuration failure events are recorded in real time and trigger alarms, supporting rapid response and fault handling. The configuration update parameters are used as output field names for this step and are subsequently called by S120's "Sensor Network Configuration" for calibrating sensor node anti-interference parameters and maintaining network stability. It also supports sensor data preprocessing and intelligent early warning model construction across main steps.

Claims

1. A method for monitoring arable land quality, characterized in that, include: The system acquires soil temperature, humidity, pH, and nutrient parameters. Using farmland geographic coordinates and soil type information as input, it performs multi-parameter sensor network deployment, adjusts transmit power, receive sensitivity, and channel selection strategies to perform anti-interference parameter calibration, including clock synchronization correction, and generates preprocessed sensor data. Based on preprocessed sensor data, reasonable threshold ranges for physical and chemical parameters are set for outlier filtering, the time zone and format of the unified timestamp are standardized to Coordinated Universal Time (UTC) for format conversion, and missing value imputation is performed using linear interpolation, spline interpolation, and machine learning-based predictive imputation to generate a complete sensor dataset. Based on a complete sensor dataset, the sliding window is segmented according to the preset sliding window size and step size parameters, an autoregressive moving average model is used for trend fitting analysis, and a symmetric encryption algorithm combined with a key management mechanism is used for encryption and compression processing to generate a cloud storage data structure. Based on the cloud storage data structure, multi-dimensional data parsing is performed according to the metadata fields in the cloud storage data structure to generate line charts, heat maps, radar charts and trend curves for dynamic chart rendering, and access control lists and role-based permission management mechanisms are used for permission verification to generate dynamic monitoring reports. The system acquires dynamic monitoring reports, uses isolated forest, support vector machine and cluster analysis methods for abnormal pattern recognition, employs multi-level threshold strategy for threshold comparison analysis and rule engine architecture for response rule matching, and generates an intelligent early warning strategy library. Based on the intelligent early warning strategy library, real-time data stream monitoring and early warning level classification are performed to generate sensor network configuration update parameters.

2. The method according to claim 1, characterized in that, The process of outlier filtering includes: Outlier filtering is performed by setting reasonable threshold ranges for physical and chemical parameters and comparing each data point collected by each sensor one by one. Based on the calculation of the mean and standard deviation of the sliding window, local abrupt changes and abnormal fluctuations are identified, and combined with historical data trends, it is determined whether the data points are abnormal.

3. The method according to claim 1, characterized in that, The process of standardizing the time zone and format of a unified timestamp to the Coordinated Universal Time (UTC) standard includes: Standardized format conversion uses a timestamp correction algorithm to adjust non-standard time formats and unify the time zone and format of timestamps to the Coordinated Universal Time standard. Data format differences between different sensor types are uniformly processed by adopting data type conversion, unit conversion, and coding standards.

4. The method according to claim 1, characterized in that, The process of using linear interpolation and spline interpolation includes: Missing value imputation is handled using a variety of imputation techniques, including linear interpolation, spline interpolation, and machine learning-based predictive imputation. For short-term missing data, linear or spline interpolation methods are preferred. For long-term missing data or complex missing data patterns, time series prediction models trained on historical data are used for filling.

5. The method according to claim 1, characterized in that, The process of segmenting the sliding window based on preset sliding window size and step parameters includes: The sliding window segmentation process divides the complete sensor dataset into time series segments based on preset sliding window size and step size parameters, forming a series of overlapping or non-overlapping time window data blocks. The size and step size of the sliding window are dynamically adjusted according to the temporal characteristics of soil parameter changes and monitoring requirements.

6. The method according to claim 1, characterized in that, The process of using an autoregressive moving average model includes: Trend fitting analysis includes performing feature extraction operations on multi-parameter data within each time window, and calculating statistical and temporal features; The variation trends of soil quality parameters were modeled using polynomial fitting, exponential smoothing, and autoregressive moving average models. Combining soil environmental factors and historical trends, the variation curves of each parameter were fitted.

7. The method according to claim 1, characterized in that, The process of performing trend fitting analysis includes: The encrypted compression process uses a symmetric encryption algorithm combined with a key management mechanism, dynamically adjusting the encryption strength and key update frequency based on the data sensitivity level and access permission policy. Lossless compression algorithms are used to efficiently compress structured time series trend data, and differential coding, entropy coding and dictionary coding are selected to optimize the compression ratio.

8. The method according to claim 2, characterized in that, The expressions for generating preprocessed sensor data include: Calculate the optimal spacing between sensor nodes: ; in, This represents the entropy value of the deployment area; A soil type classification index; Number of soil type classifications; For the first The probability percentage of soil types in the deployment area; For the first Identifiers for soil types; An improved Shannon entropy index modulator; Cross-correlation analysis is used to optimize communication frequency band selection and calculate the frequency band interference coefficient: ; in, This represents the cross-correlation value across frequency bands. and These represent the signal waveforms of the main frequency band and the interference frequency band, respectively. For delay parameters; It is a time variable; Construct the objective function: ; in, This represents the transmit power of the sensor node; This refers to the receiving sensitivity of the sensor node; Maximum permissible power; These are the weighting coefficients; This represents the cross-correlation value across frequency bands. Interference threshold; Correcting clock synchronization errors: ; in, This represents the average clock deviation. The total number of nodes; For sensor node indexing; For the first The local timestamp of each node; As a reference time base; Define data integrity metrics: ; in, For data integrity rate; The number of valid data packets; This represents the total number of data packets; This is the time deviation attenuation coefficient; This represents the average clock offset. Execution timestamp alignment: ; in, The optimal alignment reference time is calculated; The baseline time variable to be optimized; Indexing timestamp data; For the first One timestamp to be aligned; This represents the number of nodes.

9. The method according to claim 4, characterized in that, The expressions for generating cloud storage data structures include: Define window stability metrics: ; in, An improved Chebyshev norm for window data; A multi-parameter data matrix within a time window; This is the row index of the data matrix; For column indices of the data matrix; The dimension of the data matrix; The first in the data matrix row element value, The first in the data matrix The element values ​​of the column; For smoothing coefficients; It is a fixed label that indicates a specific parameter type or sensor type; Calculate window overlap ratio: ; in, The overlap rate; and These are the overlap duration and the window duration, respectively. To improve the Chebyshev norm; It is a regular Chebyshev norm; Construct the prediction equation: ; in, These are predicted values ​​for soil parameters; For time indexing; The order of the autoregressive model; and These are the coefficients for autoregression and moving average; This is the lag order index for the autoregressive term; The order of the moving average model; This is the lag order index for the moving average term; It is a white noise sequence; Calculate the trend confidence level: ; in, Confidence level; Index the data points; This represents the total number of data points within the current time window. For the first The fitting residuals of each data point; For the first The actual observed values ​​of each data point; Define the compression factor: ; in, To compress the Reynolds number; For data density; For data flow rate; This refers to the data block size. The coefficient of compressibility viscosity; For trend confidence.

10. The method according to claim 4, characterized in that, The process of generating quality change trend data includes: Feature extraction is performed on multi-parameter data within each time window, including calculating statistical features such as mean, median, variance, skewness, and kurtosis to reflect the central tendency and distribution characteristics of soil parameters; extracting temporal features such as maximum, minimum, and rate of change to describe the dynamic changes of parameters; calculating cumulative content and variation amplitude for nutrient parameters to assess the temporal evolution of soil nutrients; and combining sensor data quality labels to weight abnormal or missing data to reduce the impact of outliers on feature calculation.

Images