A smart agricultural environment monitoring system based on the Internet of Things

Abstract

The present invention provides an intelligent agricultural environment monitoring system based on the Internet of Things, which relates to the field of data processing technology. The system comprises: an environmental information acquisition module; a feature extraction and analysis module, which is used to perform segmented statistical analysis on the original soil moisture data based on the original soil moisture data, combined with historical soil moisture data, historical ambient temperature data and preset recent-period soil moisture feedback data, and automatically extract the moisture change rate and abnormal trend indicator data sets under multiple time windows; a self-learning judgment module, which is used to dynamically adjust the current judgment threshold based on a self-learning algorithm, and judge the original soil moisture data in combination with the moisture change rate and abnormal trend indicator, and generate a judgment signal whether to trigger a local irrigation response; a response control module, which is used to automatically generate an irrigation equipment startup instruction; and an irrigation execution module, which is used to execute an irrigation operation according to the irrigation equipment startup instruction. The present invention improves the accuracy of agricultural environment monitoring.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a smart agricultural environment monitoring system based on the Internet of Things. Background Art

[0002] Existing agricultural environmental monitoring systems typically deploy a variety of sensors (such as temperature and humidity sensors, light intensity sensors, and soil moisture sensors) to collect environmental data. This data is aggregated via a local gateway and uploaded to a cloud platform. System administrators can view this data via a web or mobile app and set warning thresholds, enabling remote monitoring and management of the agricultural environment. Some systems incorporate IoT communication technologies (such as LoRa and NB-IoT) to improve transmission efficiency and enhance monitoring coverage. These systems are widely used in agricultural scenarios such as greenhouses and smart irrigation.

[0003] However, in distributed farmland irrigation scenarios, existing monitoring systems are prone to delayed responses to abnormal information. For example, in water-saving irrigation systems in drought-prone areas, sensors detect that soil moisture remains persistently below a set threshold. This information must be processed by the gateway and the cloud before the irrigation system can be activated. This entire process is subject to data transmission delays and control decision lags, especially when communication quality is unstable or data is congested. This can lead to prolonged crop water shortages, severely impacting crop growth cycles and yields. Summary of the Invention

[0004] The purpose of the present invention is to provide a smart agricultural environment monitoring system based on the Internet of Things, aiming to solve the problems mentioned in the background technology.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] A smart agricultural environment monitoring system based on the Internet of Things, the system comprising:

[0007] Environmental information collection module, used to periodically collect soil moisture data from multiple monitoring points to form raw soil moisture data;

[0008] The feature extraction and analysis module is used to perform segmented statistical analysis on the original soil moisture data based on the combination of historical soil moisture data, historical ambient temperature data, and preset recent period soil moisture feedback data, and automatically extract the moisture change rate and abnormal trend indicator data sets under multiple time windows;

[0009] A self-learning judgment module is used to dynamically adjust the current judgment threshold based on the self-learning algorithm according to the moisture change rate and abnormal trend indicator data set, judge the raw soil moisture data, and generate a judgment signal on whether to trigger a local irrigation response. The initial value of the current judgment threshold is a preset value;

[0010] A response control module is used to automatically generate an irrigation equipment start instruction when the judgment signal indicates an abnormal state, otherwise continuously monitor and wait for a new judgment signal input;

[0011] The irrigation execution module is used to execute irrigation operations according to the start-up instructions of the irrigation equipment, and at the same time collect soil moisture feedback data after the irrigation is completed, and send the soil moisture feedback data and soil moisture data synchronously to the feature extraction and analysis module for it to continuously optimize the judgment threshold and judgment strategy.

[0012] Preferably, based on the original soil moisture data, combined with historical soil moisture data, historical ambient temperature data and preset recent period soil moisture feedback data, the original soil moisture data is statistically analyzed in sections to automatically extract the humidity change rate and abnormal trend indicators under multiple time windows, including:

[0013] Based on the original soil moisture data and historical soil moisture data, the original soil moisture data is aggregated in segments based on a set time window to generate aggregated data for each target time period;

[0014] Perform differential calculations on the aggregated data of each target time period and the historical soil moisture data to obtain the moisture change rate in each time window;

[0015] Combined with historical ambient temperature data and preset recent soil moisture feedback data, a multidimensional cross-statistical analysis was conducted on the humidity change rate. The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index and humidity recovery lag index after irrigation were calculated to generate an abnormal trend indicator data set.

[0016] Preferably, according to the moisture change rate and abnormal trend indicator data set, the current judgment threshold is dynamically adjusted based on the self-learning algorithm, the raw soil moisture data is judged, and a judgment signal is generated to determine whether to trigger a local irrigation response, including:

[0017] Based on the moisture change rate and abnormal trend indicator data set, the judgment score is calculated and compared with the current judgment threshold to preliminarily determine whether the current soil moisture is in a response-required state and generate a preliminary judgment signal for local irrigation response;

[0018] Collect soil moisture feedback data returned by the irrigation execution module and associate the feedback data with the preliminary judgment signal, the corresponding moisture change rate, the abnormal trend indicator data set and the judgment score to form a historical judgment sample data set;

[0019] Based on the self-learning algorithm, the moisture change rate, abnormal trend indicator data, judgment score, judgment signal and soil moisture feedback data in the historical judgment sample data set are analyzed to train the model, dynamically adjust and optimize the current judgment threshold to adapt it to soil moisture and environmental changes, and generate the optimized current judgment threshold;

[0020] Based on the optimized current judgment threshold, combined with the latest acquired moisture change rate and abnormal trend indicators, the soil moisture raw data is judged to generate the latest local irrigation response judgment signal, and the latest judgment signal and soil moisture feedback data are continuously included in the historical judgment sample data set to form an adaptive optimization closed loop of judgment thresholds and strategies.

[0021] Preferably, the aggregated data of each target time period is differentially calculated with the historical soil moisture data to obtain the moisture change rate in each time window, including:

[0022] Based on the aggregated data of each target time period, the current average soil moisture value of each target time period is extracted;

[0023] According to each target time period, the historical average soil moisture value of the corresponding time period in the historical soil moisture data is extracted;

[0024] The current average soil moisture value of each target time period is compared with its corresponding historical average soil moisture value. The relative change calculation method is used to divide the difference between the two by the historical average soil moisture value to obtain the humidity change rate in this time window.

[0025] Preferably, a multi-dimensional cross-statistical analysis of the humidity change rate is performed in combination with historical ambient temperature data and preset recent period soil moisture feedback data. The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index, and humidity recovery lag index after irrigation are calculated respectively to generate an abnormal trend indicator data set, including:

[0026] The humidity change rate in each time window is combined with historical ambient temperature data and the correlation analysis method is used to obtain the temperature-humidity correlation index;

[0027] The fluctuation amplitude and frequency of humidity change rate in each time window are statistically analyzed to generate humidity change variance index and abnormal fluctuation frequency index;

[0028] Calculate the humidity change rate index and the duration index of the abnormality according to the humidity change rate data within the preset recent period;

[0029] Conduct time series analysis on preset near-period soil moisture feedback data to determine the lag degree of moisture recovery and generate a moisture recovery lag index after irrigation.

[0030] The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index and humidity recovery lag index after irrigation are combined into one set to generate an abnormal trend index dataset.

[0031] Preferably, a judgment score is calculated based on the moisture change rate and abnormal trend indicator data set, and compared with the current judgment threshold to preliminarily judge whether the current soil moisture is in a response-required state, and generate a preliminary judgment signal for local irrigation response, including:

[0032] Normalize each indicator in the humidity change rate and abnormal trend indicator data set in each time window to make each indicator have a unified numerical range and eliminate the influence of different dimensions to obtain a normalized indicator data set;

[0033] Assign weights to the indicators in the normalized indicator dataset based on preset or self-learned weight parameters. The weight assignment can be dynamically adjusted based on historical irrigation results, enabling the judgment model to adaptively adjust the importance of different time periods and different features.

[0034] Multiply each normalized indicator value by its corresponding weight one by one, and sum all weighted results to obtain the judgment score of each time window to form a judgment score sequence;

[0035] In the judgment scoring sequence, the judgment score of each time window is compared with the current judgment threshold one by one. If the judgment score reaches or exceeds the current judgment threshold, a preliminary judgment signal of local irrigation that requires response is output, and the judgment score is recorded together with the relevant feature data for subsequent self-learning training.

[0036] Preferably, based on a self-learning algorithm, the moisture change rate, abnormal trend indicator data, judgment score, judgment signal and soil moisture feedback data in the historical judgment sample data set are analyzed to train a model, dynamically adjust and optimize the current judgment threshold to adapt it to soil moisture and environmental changes, and generate an optimized current judgment threshold, including:

[0037] For each group of samples in the historical judgment sample data set, the corresponding humidity change rate, abnormal trend index data, judgment score, preliminary judgment signal and soil moisture feedback data are extracted;

[0038] Based on the relationship between the moisture change rate, abnormal trend index data, judgment score and soil moisture feedback data of each group of samples, the actual error of each judgment is calculated, and samples with larger actual errors are given higher weights to form weighted statistical results;

[0039] According to the weighted statistical results, a self-learning algorithm is used to analyze the setting deviation of the judgment threshold, and combined with the current soil moisture raw data, the current judgment threshold is adjusted to obtain the optimized current judgment threshold.

[0040] Preferably, the humidity change rate in each time window is combined with historical ambient temperature data and a correlation analysis method is used to obtain a temperature-humidity correlation index, including:

[0041] According to the humidity change rate in each time window, the historical ambient temperature data corresponding to the time window is extracted to construct the humidity change rate time series and the ambient temperature series;

[0042] Normalize the humidity change rate time series and the ambient temperature series so that the two series have the same numerical range and eliminate the influence of different dimensions;

[0043] The normalized humidity change rate time series and the ambient temperature series are paired one by one according to the time window order, the product sum of each paired point is calculated, and the respective mean and variance are statistically calculated to obtain the mean variance data;

[0044] According to the mean-variance data, the Pearson correlation coefficient calculation method was used to solve the correlation coefficient of the two normalized series to quantify the linear correlation between the humidity change rate and the ambient temperature;

[0045] The correlation coefficient is output as the temperature and humidity correlation index.

[0046] Preferably, the humidity change rate index and the duration index of the abnormality are calculated based on the humidity change rate data within a preset recent period, including:

[0047] According to the humidity change rate data within the preset recent period, the humidity change rate time series is sorted in chronological order;

[0048] For each continuous time interval in the humidity change rate time series, calculate the numerical change of the humidity change rate within the interval, and take the length of each time interval as the benchmark, divide the change by the time interval length to obtain the change rate of each period;

[0049] The change rate of all consecutive time periods is averaged to obtain the humidity change rate index of the monitoring point in the recent period;

[0050] In the humidity change rate time series, the normal interval threshold of the humidity change rate is set to determine whether each time point is abnormal. The time period continuously in the abnormal interval is accumulated as the abnormal duration to obtain the continuous abnormal duration index.

[0051] Preferably, a time series analysis is performed on the preset recent period soil moisture feedback data to determine the hysteresis degree of humidity recovery and generate a humidity recovery hysteresis index after irrigation, including:

[0052] According to the preset near-period soil moisture feedback data, it is sorted into a humidity feedback time series according to the collection time sequence;

[0053] According to the humidity feedback time series, mark the humidity feedback value corresponding to the end time of the irrigation operation and each subsequent time;

[0054] Set the interval threshold of the normal recovery interval of soil moisture, judge the recovery of the moisture feedback value based on the interval threshold, and identify the continuous time period from the lowest moisture point at the end of irrigation to the first time it reaches the normal recovery interval;

[0055] The length of each continuous time period is taken as the humidity recovery hysteresis value and output as the humidity recovery hysteresis index after irrigation.

[0056] The above solution of the present invention includes at least the following beneficial effects:

[0057] By introducing a distributed local decision-making mechanism into the smart agricultural environmental monitoring system, this invention enables localized intelligent analysis and response to raw soil moisture data periodically collected from multiple monitoring points, effectively avoiding the decision-making delays associated with existing technologies that rely on centralized cloud-based processing. The system's feature extraction and analysis module automatically extracts moisture change rates and abnormal trend indicators across multiple time windows by combining historical soil moisture data, historical ambient temperature data, and recent soil moisture feedback data, significantly improving the scientific and targeted nature of data processing.

[0058] Compared to traditional systems that rely solely on a single threshold, the self-learning decision module in this invention dynamically adjusts the current decision threshold based on this multi-dimensional data. It also combines the latest humidity change rate and abnormal trend indicators to generate a real-time decision signal to trigger a local irrigation response. This response control module makes irrigation decisions directly locally, immediately generating irrigation equipment activation instructions when an abnormal condition is detected. This significantly reduces reliance on cloud-based transmission and remote processing, improving the real-time and reliability of the overall response.

[0059] Through this approach, the system can maintain rapid perception and timely response to farmland soil moisture conditions even in complex environments such as unstable communication quality and data congestion, reducing crop growth cycle delays and yield losses caused by water shortages. Furthermore, the irrigation feedback data collected by the irrigation execution module can feed back into the continuous optimization of the decision strategy, forming an intelligent, adaptive closed-loop control system. Overall, this invention provides technical support for large-scale distributed farmland intelligent monitoring and precision irrigation management, taking into account the system's real-time performance, adaptability, and intelligence, and possessing significant application and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is an architecture diagram of an Internet of Things-based smart agricultural environment monitoring system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0062] like Figure 1 As shown, an embodiment of the present invention provides a smart agricultural environment monitoring system based on the Internet of Things, the system comprising:

[0063] Environmental information collection module, used to periodically collect soil moisture data from multiple monitoring points to form raw soil moisture data;

[0064] The feature extraction and analysis module is used to perform segmented statistical analysis on the original soil moisture data based on the combination of historical soil moisture data, historical ambient temperature data, and preset recent period soil moisture feedback data, and automatically extract the moisture change rate and abnormal trend indicator data sets under multiple time windows;

[0065] A self-learning judgment module is used to dynamically adjust the current judgment threshold based on the self-learning algorithm according to the moisture change rate and abnormal trend indicator data set, judge the raw soil moisture data, and generate a judgment signal on whether to trigger a local irrigation response. The initial value of the current judgment threshold is a preset value;

[0066] A response control module is used to automatically generate an irrigation equipment start instruction when the judgment signal indicates an abnormal state, otherwise continuously monitor and wait for a new judgment signal input;

[0067] The irrigation execution module is used to execute irrigation operations according to the start-up instructions of the irrigation equipment, and at the same time collect soil moisture feedback data after the irrigation is completed, and send the soil moisture feedback data and soil moisture data synchronously to the feature extraction and analysis module for it to continuously optimize the judgment threshold and judgment strategy.

[0068] In an embodiment of the present invention, the system realizes the periodic collection of soil moisture data from multiple monitoring points in the agricultural planting area through the environmental information collection module. The collected data is the original soil moisture data, which provides basic information for subsequent data analysis and intelligent decision-making. The system is provided with a feature extraction and analysis module, which can systematically analyze the collected original soil moisture data. During the analysis process, historical soil moisture data, historical ambient temperature data and recent period soil moisture feedback data are combined, and the multi-dimensional and multi-time period environmental and operational information is fully utilized. The soil moisture original data is subjected to segmented statistical analysis, and the humidity change rate and multiple abnormal trend indicators under multiple time windows are automatically extracted. Finally, a data set containing the humidity change rate and abnormal trend indicators is generated, making subsequent judgments and controls more accurate and comprehensive.

[0069] In this system, the core self-learning judgment module dynamically adjusts the current judgment threshold based on extracted moisture change rate and abnormal trend indicator datasets, combined with a self-learning algorithm. The initial judgment threshold is a preset value, and the system continuously adaptively optimizes the threshold based on the actual operating environment and historical judgment results, ensuring that the judgment conditions closely match actual soil moisture and environmental changes. In actual application, the self-learning judgment module evaluates the current raw soil moisture data and automatically generates a judgment signal to determine whether a local irrigation response is necessary, greatly improving the scientific nature and accuracy of irrigation responses.

[0070] In addition, the response control module can automatically issue irrigation equipment start-up instructions based on the status of the judgment signal. When the judgment signal is abnormal, the irrigation equipment will be automatically activated, realizing unattended intelligent irrigation. When the judgment signal does not meet the start-up criteria, the system will remain in monitoring mode and wait for a new judgment signal input, avoiding ineffective or excessive irrigation and ensuring the effective use of water resources. The irrigation execution module is responsible for executing the actual irrigation operations according to the instructions of the control module, and simultaneously collecting soil moisture feedback data after irrigation. The latest feedback data is promptly transmitted back to the feature extraction and analysis module, providing the latest basis for the system to adaptively adjust the judgment threshold and strategy. This closed-loop optimization mechanism not only enhances the intelligence level of monitoring and control, but also ensures the adaptability and stability of the system's long-term operation, significantly improving the production efficiency and water management of smart agriculture.

[0071] The response control module is used to automatically generate an irrigation equipment start instruction when the judgment signal indicates an abnormal state, otherwise continuously monitor and wait for new judgment signal input, specifically including:

[0072] After receiving the local irrigation response determination signal from the learning determination module, the response control module first identifies and determines the content of the determination signal. If the determination signal indicates "response required" or indicates an abnormal state, the module immediately generates an irrigation equipment start-up instruction and transmits this instruction to the corresponding irrigation execution module via a wireless communication module or a wired network to ensure that farmland irrigation can be initiated in the shortest possible time. For example, in actual applications, when the system continuously detects that the soil moisture in a certain area has been below the optimized current determination threshold for a long period of time, and the moisture change rate and abnormal trend indicators are combined to determine that irrigation is required, the response control module will issue a specific irrigation start-up command, clearly specifying the irrigation equipment number, start-up duration, or water level to be activated, to ensure that the soil moisture returns to a reasonable range as soon as possible.

[0073] If the judgment signal fails to meet the abnormal response conditions, meaning the system deems the current soil moisture in the farmland to be within the normal range, the response control module will not issue the irrigation equipment start command. Instead, it will maintain monitoring status, continuously receiving and waiting for the next judgment signal input. Based on the configured sampling period, scheduled polling, or event triggering, the module continuously receives the latest data from the environmental information acquisition module and feature extraction and analysis module, re-triggering the intelligent judgment of the self-learning judgment module to ensure real-time and continuous system response. For example, if the sampling period is 15 minutes, the system will update the data every 15 minutes, automatically re-evaluating the soil condition and maintaining dynamic tracking of the farmland environment.

[0074] To prevent false triggering due to abnormal conditions or prolonged periods of inactivity, the response control module can also incorporate a multi-point decision-making mechanism or a redundant verification mechanism for judgment signals. For example, the module can set a rule requiring two or more consecutive abnormal judgment signals before issuing an irrigation start instruction. Alternatively, the module can integrate data from different monitoring points to perform a multi-point weighted comprehensive judgment, thereby improving the reliability of the system's judgment. In practice, if a monitoring point briefly falsely alarms while overall soil moisture conditions are good, the system can avoid false irrigation, thereby saving water and extending equipment life.

[0075] The response control module also supports priority management and device linkage in abnormal conditions. For example, when the system detects abnormality signals simultaneously at different monitoring points, it can prioritize irrigation instructions for key areas based on regional importance, historical irrigation water usage, and crop growth stage, while dynamically adjusting response strategies for other areas to ensure rational resource allocation and efficient system response.

[0076] In summary, the response control module achieves efficient and automatic control of agricultural irrigation equipment through intelligent analysis of judgment signals, command generation, continuous monitoring, and multi-point redundancy verification. This improves the intelligent response capabilities and practical reliability of the smart agricultural environmental monitoring system. This module's simple structure and open interface facilitate integration with mainstream agricultural IoT platforms, meeting the needs of automated irrigation in diverse farmland scenarios.

[0077] In a preferred embodiment of the present invention, based on the original soil moisture data, combined with historical soil moisture data, historical ambient temperature data and preset recent period soil moisture feedback data, the original soil moisture data is statistically analyzed in sections to automatically extract the moisture change rate and abnormal trend indicators under multiple time windows, including:

[0078] Based on the original soil moisture data and historical soil moisture data, the original soil moisture data is aggregated in segments based on a set time window to generate aggregated data for each target time period;

[0079] Perform differential calculations on the aggregated data of each target time period and the historical soil moisture data to obtain the moisture change rate in each time window;

[0080] Combined with historical ambient temperature data and preset recent soil moisture feedback data, a multidimensional cross-statistical analysis was conducted on the humidity change rate. The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index and humidity recovery lag index after irrigation were calculated to generate an abnormal trend indicator data set.

[0081] In an embodiment of the present invention, the system can carry out multi-dimensional segmented statistical analysis based on the original soil moisture data, combined with historical soil moisture data, historical ambient temperature data and preset near-period soil moisture feedback data, to achieve automatic extraction of humidity change rate and abnormal trend indicators. During the specific operation, first, the system segments and aggregates the original soil moisture data according to the set time window based on the original soil moisture data and historical soil moisture data, and generates the current average soil moisture value for each target time period. Subsequently, for each target time period, the system extracts the historical average soil moisture value of the corresponding time period in the historical soil moisture data, and uses the relative change method to divide the difference between the current average value and the historical average value by the historical average value to accurately obtain the humidity change rate under each time window. This processing method can fully reflect the dynamic changes in the crop growth environment, avoid misjudgment due to differences in absolute values, and improve the environmental adaptability of the system.

[0082] The system also performs multi-dimensional cross-statistical analysis of humidity change rate, historical ambient temperature data, and pre-set recent soil moisture feedback data. Specifically, the humidity change rate within each time window is combined with historical ambient temperature data, and correlation analysis is used to quantify the correlation between humidity change and temperature, resulting in a temperature-humidity correlation index. Statistical analysis of the humidity change rate time series can also be performed on the fluctuation amplitude and frequency, yielding humidity change variance and abnormal fluctuation frequency indices, respectively, to accurately identify abnormal phenomena such as short-term, drastic changes or long-term fluctuations. For the humidity change rate index and duration of abnormality index, the system segments and applies thresholds to the humidity change rate series within a pre-set recent period to accurately capture humidity change trends and their persistence. For the humidity recovery lag after irrigation index, the system uses time series analysis of recent soil moisture feedback data to determine the time required for humidity to return to the normal range from its lowest point, scientifically reflecting the soil's recovery rate after irrigation.

[0083] Ultimately, all analysis results are integrated into a dataset of abnormal trend indicators, ensuring the system can make highly accurate and personalized responses to crop growth conditions in diverse environments and historical periods. This multi-layered, multi-dimensional indicator system significantly improves the sensitivity and accuracy of environmental monitoring, providing solid data support for subsequent intelligent irrigation control and decision model optimization.

[0084] Among them, according to the original soil moisture data and historical soil moisture data, the original soil moisture data is segmented and aggregated based on the set time window to generate aggregated data for each target time period, specifically including:

[0085] The system first acquires the raw soil moisture data collected regularly by the environmental information acquisition module and simultaneously retrieves historical soil moisture data corresponding to the current monitoring period. To facilitate analysis and comparison, the system pre-sets appropriate time windows, such as 5 minutes, 30 minutes, 1 hour, and 24 hours. The specific window length can be flexibly configured based on crop type, growth stage, and soil characteristics. Each time window covers a continuous sampling period.

[0086] In practice, the system groups all raw soil moisture data collected from a monitoring point within a 24-hour period into hourly time windows. For example, 8:00 a.m. to 9:00 a.m. is considered a time window, and the system collects and aggregates soil moisture data from all sampling points within that time period. For each time window, the system aggregates the raw soil moisture data within the group using statistical methods such as mean, maximum, minimum, and median to obtain aggregated data for the current time window. Taking the mean as an example, the system sums all the data within the time period and divides it by the number of data points to obtain a representative value reflecting the soil moisture level during that time period.

[0087] The system also processes historical soil moisture data using the same time segmentation rules as the current time window to ensure comparability across different time periods. For example, when comparing data across years, seasons, or the same period in history, the system extracts historical sampling data from the same period of the previous year, quarter, or month and aggregates them within the same time window to facilitate subsequent difference analysis and trend determination.

[0088] If monitoring points are widely distributed or the data volume is large, the system can use sliding time window technology to achieve continuous coverage. This technology generates new time windows at set sliding steps (e.g., 15 minutes) to improve data resolution. For example, if the window length is set to 1 hour and the sliding step is 15 minutes, the system will update the window every 15 minutes, aggregating soil moisture data from the previous hour and effectively capturing short-term fluctuations.

[0089] For time windows with missing data, the system can use interpolation, historical averaging, or intelligent estimation to complete the data. For example, if some sample data is missing within a time window, the system can replace it with the average value of the monitoring point within the most recent same time window, ensuring the continuity and integrity of data analysis.

[0090] Through the above-mentioned segmented aggregation method, the system can efficiently and scientifically convert large quantities of original collected data into structured aggregated data, laying a solid foundation for subsequent complex statistical analysis such as humidity change rate and trend indicators, enabling technical personnel in this field to easily implement the data segmentation and aggregation operations of the present invention based on this description, and it is easy to flexibly expand and optimize in different agricultural scenarios.

[0091] In a preferred embodiment of the present invention, based on the moisture change rate and abnormal trend indicator data set, the current judgment threshold is dynamically adjusted based on the self-learning algorithm, the raw soil moisture data is judged, and a judgment signal is generated to determine whether to trigger a local irrigation response, including:

[0092] Based on the moisture change rate and abnormal trend indicator data set, the judgment score is calculated and compared with the current judgment threshold to preliminarily determine whether the current soil moisture is in a response-required state and generate a preliminary judgment signal for local irrigation response;

[0093] Collect soil moisture feedback data returned by the irrigation execution module and associate the feedback data with the preliminary judgment signal, the corresponding moisture change rate, the abnormal trend indicator data set and the judgment score to form a historical judgment sample data set;

[0094] Based on the self-learning algorithm, the moisture change rate, abnormal trend indicator data, judgment score, judgment signal and soil moisture feedback data in the historical judgment sample data set are analyzed to train the model, dynamically adjust and optimize the current judgment threshold to adapt it to soil moisture and environmental changes, and generate the optimized current judgment threshold;

[0095] Based on the optimized current judgment threshold, combined with the latest acquired moisture change rate and abnormal trend indicators, the soil moisture raw data is judged to generate the latest local irrigation response judgment signal, and the latest judgment signal and soil moisture feedback data are continuously included in the historical judgment sample data set to form an adaptive optimization closed loop of judgment thresholds and strategies.

[0096] In an embodiment of the present invention, the system can dynamically adjust the current judgment threshold based on the humidity change rate and abnormal trend indicator dataset, combined with a self-learning algorithm, to achieve intelligent judgment of the raw soil moisture data and automatically generate a judgment signal to trigger a local irrigation response. In the specific implementation process, the system will first comprehensively calculate the judgment score for each time window based on the humidity change rate and abnormal trend indicator dataset of the latest cycle through normalization, weighted scoring and other models. Subsequently, the judgment score is compared with the current judgment threshold in real time. If the score reaches or exceeds the threshold, the system automatically outputs a preliminary judgment signal for the local irrigation response, achieving an immediate response to the irrigation requirement status. This judgment process fully considers multiple dynamic environmental characteristics, avoids misjudgments caused by a single threshold, and improves the accuracy and reliability of the irrigation response.

[0097] The system also automatically collects soil moisture feedback data returned by the irrigation execution module and associates the feedback data with the generated preliminary judgment signal, the corresponding moisture change rate, the abnormal trend indicator data set, and the judgment score to form a historical judgment sample data set. This sample data set is continuously collected and updated by the self-learning judgment module to ensure that the system can fully record the true correspondence between each judgment and the actual feedback. Using a self-learning algorithm, the system jointly analyzes the moisture change rate, abnormal trend indicator data, judgment score, judgment signal, and soil moisture feedback data in the historical judgment sample data set, and dynamically optimizes the current judgment threshold through model training. For each judgment result with a large error, the system can assign a higher weight to achieve adaptive learning and strategy correction for extreme or abnormal environments. Each optimized judgment threshold will be applied in real time to the latest data judgment and incorporated into the historical judgment sample data set together with the continuously collected soil moisture feedback data, further enhancing the effectiveness of the system's adaptive optimization closed loop.

[0098] Through this design, the system continuously adjusts and optimizes its criteria for different environments and crop growth stages, ensuring that each irrigation response is both scientific and efficient. This continuous optimization and dynamic learning mechanism ensures the system maintains a high level of adaptability and intelligence over the long term, effectively supporting high-quality, energy-efficient automated irrigation management for smart agriculture.

[0099] In a preferred embodiment of the present invention, a differential operation is performed on the aggregated data of each target time period and the historical soil moisture data to obtain the moisture change rate in each time window, including:

[0100] Based on the aggregated data of each target time period, the current average soil moisture value of each target time period is extracted;

[0101] According to each target time period, the historical average soil moisture value of the corresponding time period in the historical soil moisture data is extracted;

[0102] The current average soil moisture value of each target time period is compared with its corresponding historical average soil moisture value. The relative change calculation method is used to divide the difference between the two by the historical average soil moisture value to obtain the humidity change rate in this time window.

[0103] In this embodiment of the present invention, the system extracts the current average soil moisture value for each target time period based on the aggregated data for that time period. Furthermore, the system extracts the historical average soil moisture value corresponding to each target time period from the historical soil moisture data. Both types of data use the same aggregation and time windowing rules, ensuring rigorous statistical analysis and accurate comparison.

[0104] Next, the system compares the current average soil moisture value for each target time period with the corresponding historical average soil moisture value. By calculating the relative change—dividing the difference between the two by the historical average soil moisture value—the system calculates the rate of change of moisture within each time window. This process no longer depends on absolute humidity levels, but instead dynamically reflects the current period's soil moisture relative to historical norms, allowing for timely detection of soil moisture shortages, surpluses, or abnormal fluctuations.

[0105] This embodiment achieves high-precision dynamic perception of soil moisture changes under different growth cycles and different meteorological years, providing strong basic data support for subsequent feature analysis, judgment threshold adjustment, and irrigation strategy optimization, effectively improving the scientific nature and reliability of environmental monitoring.

[0106] In a preferred embodiment of the present invention, a multi-dimensional cross-statistical analysis is performed on the humidity change rate in combination with historical ambient temperature data and preset recent period soil moisture feedback data. The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index, and humidity recovery lag index after irrigation are calculated respectively to generate an abnormal trend index data set, including:

[0107] The humidity change rate in each time window is combined with historical ambient temperature data and the correlation analysis method is used to obtain the temperature-humidity correlation index;

[0108] The fluctuation amplitude and frequency of humidity change rate in each time window are statistically analyzed to generate humidity change variance index and abnormal fluctuation frequency index;

[0109] Calculate the humidity change rate index and the duration index of the abnormality according to the humidity change rate data within the preset recent period;

[0110] Conduct time series analysis on preset near-period soil moisture feedback data to determine the lag degree of moisture recovery and generate a moisture recovery lag index after irrigation.

[0111] The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index and humidity recovery lag index after irrigation are combined into one set to generate an abnormal trend index dataset.

[0112] In an embodiment of the present invention, the system can conduct multi-dimensional cross-statistical analysis on the obtained humidity change rate data in combination with historical ambient temperature data and preset near-period soil moisture feedback data, and calculate the humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index and humidity recovery lag index after irrigation, thereby automatically generating an abnormal trend indicator data set.

[0113] First, the system extracts historical ambient temperature data based on the humidity change rate within each time window. Using correlation analysis, it quantitatively analyzes the relationship between humidity change rate and temperature, outputting the results as a temperature-humidity correlation index. This approach can identify the impact of ambient temperature fluctuations on soil moisture changes, helping to identify water stress risks caused by extreme weather conditions such as high temperature and low humidity in advance.

[0114] Next, the system performs statistical analysis on the time series of humidity change rates, calculating the amplitude and frequency of fluctuations in each time window. This results in an index of humidity change variance and an index of abnormal fluctuation frequency. This statistical result reveals the stability and abruptness of soil moisture changes, providing practical implications for early warning of abnormal weather and equipment failures.

[0115] Furthermore, based on pre-set, recent periodic humidity change rate data, the system identifies persistent periods of abnormal soil moisture and calculates both the rate of change and the duration of the abnormality. These indicators comprehensively reflect the persistence of soil drought or waterlogging, providing a scientific basis for precise irrigation and field management decisions in agricultural production.

[0116] Finally, the system performs a time-series analysis of recent soil moisture feedback data to identify the time required for soil moisture to return to normal after irrigation. This time is used as the post-irrigation moisture recovery hysteresis indicator. This not only reflects soil permeability and the water absorption capacity of crop roots, but can also be used to assess the water recovery characteristics of different plots or crops. All analyzed indicators are integrated into the abnormal trend indicator dataset, giving the system multi-dimensional intelligent perception and adaptive judgment capabilities, significantly improving the comprehensive monitoring and control capabilities of complex field environments.

[0117] In a preferred embodiment of the present invention, the humidity change rate in each time window is combined with historical ambient temperature data to obtain a temperature-humidity correlation index using a correlation analysis method, including:

[0118] According to the humidity change rate in each time window, the historical ambient temperature data corresponding to the time window is extracted to construct the humidity change rate time series and the ambient temperature series;

[0119] Normalize the humidity change rate time series and the ambient temperature series so that the two series have the same numerical range and eliminate the influence of different dimensions;

[0120] The normalized humidity change rate time series and the ambient temperature series are paired one by one according to the time window order, the product sum of each paired point is calculated, and the respective mean and variance are statistically calculated to obtain the mean variance data;

[0121] According to the mean-variance data, the Pearson correlation coefficient calculation method was used to solve the correlation coefficient of the two normalized series to quantify the linear correlation between the humidity change rate and the ambient temperature;

[0122] The correlation coefficient is output as the temperature and humidity correlation index.

[0123] In this embodiment of the present invention, the system uses correlation analysis to scientifically quantify the linear relationship between temperature and humidity by combining the humidity change rate within each time window with historical ambient temperature data, thereby obtaining a temperature-humidity correlation index. The system first groups all collected humidity change rates according to time windows and synchronously extracts the historical ambient temperature data within each corresponding time window to ensure data time alignment and matching.

[0124] On this basis, the system normalizes the humidity change rate and ambient temperature time series, aligning their values ​​within the same dimension and eliminating the effects of varying data dimensions and fluctuations. The normalized data series are then paired one by one within the time window. The system calculates the product of each paired point and calculates the mean and variance of the two series, laying the foundation for subsequent correlation calculations.

[0125] The system then systematically evaluated the strength of the correlation between the rate of humidity change and ambient temperature using the Pearson correlation coefficient analysis method, based on mean and variance results. The resulting correlation coefficient was directly used as an indicator of the temperature-humidity correlation and incorporated into the abnormal trend indicator dataset. This metric quantifies the direct impact of ambient temperature changes on soil moisture dynamics, effectively assisting in identifying water imbalance risks in special agricultural environments, such as those characterized by high temperature and high evaporation, or low temperature and high humidity. This provides strong data support for subsequent intelligent decision-making and model optimization.

[0126] In a preferred embodiment of the present invention, the humidity change rate index and the duration index of the abnormality are calculated based on the humidity change rate data within a preset recent period, including:

[0127] According to the humidity change rate data within the preset recent period, the humidity change rate time series is sorted in chronological order;

[0128] For each continuous time interval in the humidity change rate time series, calculate the numerical change of the humidity change rate within the interval, and take the length of each time interval as the benchmark, divide the change by the time interval length to obtain the change rate of each period;

[0129] The change rate of all consecutive time periods is averaged to obtain the humidity change rate index of the monitoring point in the recent period;

[0130] In the humidity change rate time series, the normal interval threshold of the humidity change rate is set to determine whether each time point is abnormal. The time period continuously in the abnormal interval is accumulated as the abnormal duration to obtain the continuous abnormal duration index.

[0131] In this embodiment of the present invention, the system can scientifically calculate a humidity change rate index and a duration of abnormality index based on pre-set humidity change rate data within a recent period, comprehensively quantifying the dynamic characteristics of soil moisture under both short-term and continuous abnormal conditions. The system first organizes the humidity change rate data collected within a recent period into a humidity change rate time series in chronological order to ensure data integrity and continuity.

[0132] For each consecutive time interval, the system calculates the numerical change in the corresponding humidity rate of change and, based on the length of the time interval, calculates the humidity change rate for that period. After calculating the change rates for all consecutive periods, the system statistically averages these change rates to ultimately determine the overall humidity change rate for the monitoring point over the recent period. This indicator reflects the rate of water loss and replenishment in the current environment or crop stage, helping to timely adjust irrigation frequency and water volume.

[0133] Based on this, the system sets reasonable thresholds for the humidity change rate within a normal range, traverses the humidity change rate time series, and determines whether anomalies occur at each time point. For time points within the abnormal range, the system automatically checks their continuity, accumulates all consecutive time periods within the abnormal range, and ultimately generates an indicator of the duration of the continuous abnormality. By accurately identifying the continuous abnormality period, the system can provide managers with early warnings of water stress, effectively preventing crop damage and achieving precise and efficient intelligent farmland environmental management.

[0134] In a preferred embodiment of the present invention, the fluctuation amplitude and frequency of the humidity change rate are statistically analyzed within each time window to generate a humidity change variance index and an abnormal fluctuation frequency index, including:

[0135] The humidity change rate values ​​of all time windows in the humidity change rate time series are calculated and their mean is used as the benchmark. The deviation between the humidity change rate of each time window and the mean is calculated. The sum of the squares of all deviations is obtained and divided by the number of time windows to obtain the humidity change variance index.

[0136] In the humidity change rate time series, the abnormal judgment threshold of the humidity change rate is set, and each time window is traversed. The number of time windows in which the humidity change rate value exceeds the abnormal judgment threshold is counted and used as the abnormal fluctuation frequency indicator.

[0137] In this embodiment of the present invention, the system can calculate the fluctuation amplitude and frequency of the humidity change rate within each time window, thereby generating a humidity change variance index and an abnormal fluctuation frequency index. The system's specific statistical process accurately depicts the dynamic changes in farmland soil moisture. The system first summarizes the humidity change rate values ​​for all time windows within the humidity change rate time series, calculates the average, and uses this average as the reference point.

[0138] The system then calculates the difference between the humidity change rate and the mean for each time window, calculating the sum of squared deviations across all time windows. This sum is then divided by the number of time windows to produce the humidity change variance index for the current period. The humidity change variance index reflects the degree of dispersion in soil moisture changes within the monitored area. A larger variance indicates more dramatic soil moisture fluctuations, helping to identify risks such as extreme weather, uneven irrigation, or improper management.

[0139] The system also sets an abnormality threshold for the humidity change rate. It then traverses the humidity change rate time series, counting the number of time windows exceeding this threshold and using this number as an abnormal fluctuation frequency indicator. This indicator reflects the frequency of abnormal soil moisture fluctuations within a certain period, providing direct warning of local environmental imbalances, sensor failures, or emergencies. By combining the humidity change variance indicator with the abnormal fluctuation frequency indicator, the system can provide farm managers with highly timely and high-resolution anomaly monitoring, effectively improving the scientific nature and responsiveness of irrigation decisions.

[0140] In a preferred embodiment of the present invention, a time series analysis is performed on preset near-period soil moisture feedback data to determine the hysteresis degree of humidity recovery and generate a post-irrigation humidity recovery hysteresis index, including:

[0141] According to the preset near-period soil moisture feedback data, it is sorted into a humidity feedback time series according to the collection time sequence;

[0142] According to the humidity feedback time series, mark the humidity feedback value corresponding to the end time of the irrigation operation and each subsequent time;

[0143] Set the interval threshold of the normal recovery interval of soil moisture, judge the recovery of the moisture feedback value based on the interval threshold, and identify the continuous time period from the lowest moisture point at the end of irrigation to the first time it reaches the normal recovery interval;

[0144] The length of each continuous time period is taken as the humidity recovery hysteresis value and output as the humidity recovery hysteresis index after irrigation.

[0145] In this embodiment of the present invention, the system performs time-series analysis on pre-set, near-period soil moisture feedback data to determine the degree of humidity recovery lag and generate a post-irrigation humidity recovery lag indicator, providing a key decision-making basis for agricultural irrigation management. The system first organizes the collected soil moisture feedback data into a humidity feedback time series sequence based on the chronological order of collection, ensuring data continuity and the integrity of the timeline.

[0146] The system accurately marks the end of each irrigation operation within this time series and locks the moisture feedback value corresponding to each time point after irrigation. To assess soil moisture recovery, the system sets thresholds for normal soil moisture recovery based on agronomic requirements or crop growth characteristics. For example, a target moisture level or range can be set as the standard for soil moisture after irrigation.

[0147] Next, the system traverses the post-irrigation humidity feedback time series, searching for the time point at which the humidity first reaches or exceeds the recovery threshold. Starting from the irrigation end point, the system identifies the continuous time period from that point until the humidity first enters the recovery range. The actual length of this time period is determined as the humidity recovery hysteresis value. A larger value indicates a more significant delay in the soil's response to irrigation, potentially indicating issues such as poor infiltration, soil compaction, or root absorption impairment.

[0148] Finally, the system outputs the humidity recovery hysteresis value as the humidity recovery hysteresis indicator after irrigation and incorporates it into the abnormal trend indicator data set. This not only provides refined feedback for irrigation decisions and system self-learning optimization, but also provides important data reference for farm managers' subsequent farmland improvements and precise operation and maintenance.

[0149] In a preferred embodiment of the present invention, a judgment score is calculated based on the moisture change rate and abnormal trend indicator data set, and compared with the current judgment threshold to preliminarily determine whether the current soil moisture is in a response-required state, and generate a preliminary judgment signal for local irrigation response, including:

[0150] Normalize each indicator in the humidity change rate and abnormal trend indicator data set in each time window to make each indicator have a unified numerical range and eliminate the influence of different dimensions to obtain a normalized indicator data set;

[0151] Assign weights to the indicators in the normalized indicator dataset based on preset or self-learned weight parameters. The weight assignment can be dynamically adjusted based on historical irrigation results, enabling the judgment model to adaptively adjust the importance of different time periods and different features.

[0152] Multiply each normalized indicator value by its corresponding weight one by one, and sum all weighted results to obtain the judgment score of each time window to form a judgment score sequence;

[0153] In the judgment scoring sequence, the judgment score of each time window is compared with the current judgment threshold one by one. If the judgment score reaches or exceeds the current judgment threshold, a preliminary judgment signal of local irrigation that requires response is output, and the judgment score is recorded together with the relevant feature data for subsequent self-learning training.

[0154] In an embodiment of the present invention, the system can comprehensively and scientifically calculate the judgment score based on the moisture change rate and abnormal trend indicator data set, combined with normalization processing and weighted scoring model, and determine whether the current soil moisture is in a response-required state based on this, thereby achieving high-precision irrigation response judgment.

[0155] Specifically, the system first normalizes each indicator in the humidity change rate and abnormal trend indicator dataset within each time window to ensure that all indicators are within a unified numerical range, eliminating the impact of dimensional or scale differences in the data itself. This normalization step ensures scientific comparability between indicators from different sources and types.

[0156] Next, the system assigns weights to each indicator in the normalized indicator dataset based on pre-set weighting parameters or dynamic adjustments from a self-learning model. The system dynamically adjusts weighting parameters based on historical irrigation performance to adapt to different time periods, soil, and climate characteristics, enabling adaptive sensitivity adjustments to key factors in the scoring model.

[0157] The system then multiplies each normalized indicator by its corresponding weight and sums all weighted results to obtain a judgment score for each time window, ultimately forming a judgment score sequence. The system traverses the judgment score sequence, comparing each score with the current judgment threshold. If the score reaches or exceeds the current threshold, it outputs a preliminary judgment signal for local irrigation response.

[0158] At the same time, the system automatically records each judgment score and its associated feature data, providing real, detailed historical sample support for subsequent self-learning judgment threshold optimization and model training. This intelligent scoring and judgment mechanism not only significantly improves judgment accuracy and robustness, but also lays a solid data foundation for the system's subsequent self-learning and adaptive evolution, effectively ensuring the long-term, efficient and stable operation of the smart agricultural environmental monitoring system.

[0159] Among them, weights are assigned to indicators in the normalized indicator dataset based on preset or self-learned weight parameters. The weight assignment can be dynamically adjusted based on historical irrigation effects to achieve adaptive adjustment of the importance of the judgment model for different time periods and different characteristics. Specifically, it includes:

[0160] In this embodiment of the present invention, the system first establishes a set of weighting parameters for each indicator within each time window, including the normalized humidity change rate, humidity change variance, frequency of abnormal fluctuations, duration of persistent abnormalities, temperature-humidity correlation, and hysteresis of humidity recovery after irrigation. These weighting parameters can initially be preset by the system, for example, by agricultural experts or managers assigning initial weights based on empirical knowledge, or by setting initial ratios based on historical data from crop types and planting regions. For example, in dryland farmland, the system might set a weight of 0.4 for the humidity change rate and 0.3 for the duration of persistent abnormalities. The remaining weights for the remaining indicators can then be assigned based on actual needs.

[0161] In actual operation, the system will continuously collect moisture feedback data and historical judgment results after each irrigation, and use these historical irrigation effects as the basis for weight optimization. Specifically, the system evaluates the prediction accuracy of each indicator by looking back at historical irrigation decisions and actual soil moisture improvements. When a certain indicator contributes more to the accurate judgment of the soil's irrigation status within a specific period, the system automatically increases its weight; if a certain indicator fails to effectively improve the prediction accuracy in multiple historical judgments, the system appropriately reduces its weight. This dynamic adjustment process can be achieved through weighted averaging, Bayesian updating, incremental learning and other methods, but all calculation results are presented in textual descriptions. For example: the system will perform statistical analysis on the corresponding errors when each indicator participates in decision-making, and automatically increase or decrease the weight according to the size of the error.

[0162] For example, assuming that the duration of anomalies over a period of time effectively predicts drought across multiple decision-making processes, while the frequency of abnormal fluctuations only provides limited improvement in accuracy, the system will analyze these historical samples and automatically increase the weight of the duration of anomalies from 0.3 to 0.35, while simultaneously decreasing the weight of the frequency of abnormal fluctuations from 0.15 to 0.1. This allows the system's judgment model to automatically adapt the importance of each feature within the model to different seasons, crop growth cycles, and irrigation response, achieving adaptive optimization.

[0163] Furthermore, dynamic adjustment of weight parameters isn't limited to global cycles; local optimization can also be performed for different time periods, weather conditions, or crop growth stages. For example, during the rainy season, the weight of temperature-humidity correlation indicators might increase, while during the dry season, the weights of the duration of abnormality and the rate of humidity change would be increased accordingly. The system can periodically and automatically restructure the distribution of weight parameters based on environmental changes and actual feedback, ensuring that each indicator maximizes its predictive value based on the actual scenario.

[0164] In a preferred embodiment of the present invention, based on a self-learning algorithm, the moisture change rate, abnormal trend indicator data, judgment score, judgment signal and soil moisture feedback data in the historical judgment sample data set are analyzed to train a model, dynamically adjust and optimize the current judgment threshold to adapt it to soil moisture and environmental changes, and generate an optimized current judgment threshold, including:

[0165] For each group of samples in the historical judgment sample data set, the corresponding humidity change rate, abnormal trend index data, judgment score, preliminary judgment signal and soil moisture feedback data are extracted;

[0166] Based on the relationship between the moisture change rate, abnormal trend index data, judgment score and soil moisture feedback data of each group of samples, the actual error of each judgment is calculated, and samples with larger actual errors are given higher weights to form weighted statistical results;

[0167] According to the weighted statistical results, a self-learning algorithm is used to analyze the setting deviation of the judgment threshold, and combined with the current soil moisture raw data, the current judgment threshold is adjusted to obtain the optimized current judgment threshold.

[0168] In this embodiment of the present invention, the system uses a self-learning algorithm to analyze the moisture change rate, abnormal trend indicator data, judgment score, judgment signal, and soil moisture feedback data in historical judgment sample data sets. This automatically trains the model, dynamically adjusts and optimizes the current judgment threshold to better adapt to the changing characteristics of soil moisture and the environment, and continuously generates optimized current judgment thresholds. For each irrigation judgment event, the system first fully records the relevant moisture change rate, abnormal trend indicator data, actual output judgment score, preliminary judgment signal, and corresponding soil moisture feedback data after irrigation completion, and organizes this information into a set of samples.

[0169] The system then performs a statistical analysis of the relationship between the above data for each set of samples, specifically analyzing the deviation between the initial judgment signal and the subsequent actual soil moisture feedback data. For example, when the initial judgment signal indicates that irrigation is needed, but the feedback data shows that the irrigation effect is not as expected, the system will classify such events as high-error samples. By assigning higher weights to these high-error samples, the system can focus on anomalies or incorrect judgments, increasing the model's sensitivity to edge cases and special circumstances. This weighted statistical result not only accelerates the model's convergence process, but also effectively prevents the threshold from being excessively affected by abnormal samples.

[0170] Based on statistical analysis results, the system uses self-learning algorithms (such as incremental learning and online regression) to analyze deviations from the current judgment threshold settings. It then automatically adjusts and optimizes the thresholds in real time, integrating newly collected raw soil moisture data, moisture change rates, and abnormal trend indicator data. The optimized current judgment thresholds are immediately applied to subsequent irrigation decision processes, ensuring that judgment results consistently adapt to dynamic changes in the environment and crop conditions. Ultimately, the system achieves continuous self-optimization of thresholds and adaptive evolution of judgment criteria, providing a highly intelligent and precise water regulation solution for agricultural production.

[0171] Among them, according to the weighted statistical results, a self-learning algorithm is used to analyze the setting deviation of the judgment threshold, and combined with the current soil moisture raw data, the current judgment threshold is adjusted to obtain the optimized current judgment threshold, specifically including:

[0172] In this embodiment of the present invention, after collecting multiple sets of historical judgment samples, the system conducts a comprehensive analysis of the moisture change rate, abnormal trend indicators, judgment scores, preliminary judgment signals, and soil moisture feedback data contained in each set of samples. The system first summarizes the errors in each judgment by statistically analyzing the difference between the judgment score and the actual irrigation effect for each set of samples. For example, if the system outputs a preliminary judgment signal of "irrigation required," but the soil moisture feedback after irrigation shows that the expected moisture improvement has not been achieved, this is considered a false positive. Samples with larger errors are given a higher weight in subsequent statistical analysis, allowing them to have a greater impact on threshold optimization.

[0173] Based on this, the system uses self-learning algorithms (such as incremental threshold adjustment strategies, sliding window averaging, or decision tree-based parameter optimization mechanisms) to analyze the distribution of judgment errors for all high-weighted samples. By analyzing the central trend and fluctuation range of these error distributions, the system determines whether the current judgment threshold is too high or too low. For example, if false positives or negatives are frequently observed in recent high-weighted samples, the system will automatically determine that the current threshold setting is biased.

[0174] The system then calibrates the current environmental conditions based on the raw soil moisture data collected during the most recent cycle, determining the magnitude and direction of the threshold adjustment. For example, if the current soil moisture is generally low, and the system error samples indicate that the current threshold is too high, resulting in multiple missed detections, the system will appropriately lower the current threshold. This adjustment can be done using a step-by-step approach, fine-tuning the threshold each time based on the average error of recent high-weighted samples until the false positive rate falls within an acceptable range.

[0175] Ultimately, through the aforementioned statistical and self-learning optimization process, the system adjusts and outputs the optimal threshold value for current environmental conditions in real time, achieving continuous self-evolution of the judgment criteria. For example, if, during a period of high summer temperatures and drought, actual feedback indicates that historical threshold settings were too high, resulting in a delayed irrigation response, the system will continuously lower the threshold, enabling the monitoring system to detect irrigation needs earlier, thereby improving the scientific nature and timeliness of irrigation management. Conversely, during rainy seasons or periods of high humidity, the system can automatically raise the threshold to avoid unnecessary irrigation operations.

[0176] Through the above method, the embodiment of the present invention enables technical personnel in this field to dynamically and accurately adjust and optimize the judgment threshold according to weighted statistical results, historical judgment data and current soil moisture original data, thereby realizing efficient adaptive control of the smart agricultural environment monitoring system.

[0177] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A smart agricultural environment monitoring system based on the Internet of Things, characterized by: The system comprises: Environmental information collection module, used to periodically collect soil moisture data from multiple monitoring points to form raw soil moisture data; The feature extraction and analysis module is used to perform segmented statistical analysis on the original soil moisture data based on the combination of historical soil moisture data, historical ambient temperature data, and preset recent period soil moisture feedback data, and automatically extract the moisture change rate and abnormal trend indicator data sets under multiple time windows; The self-learning judgment module is used to dynamically adjust the current judgment threshold based on the self-learning algorithm according to the moisture change rate and abnormal trend indicator data set, judge the original soil moisture data, and generate a judgment signal on whether to trigger a local irrigation response. The initial value of the current judgment threshold is a preset value, specifically including: Based on the moisture change rate and abnormal trend indicator data set, the judgment score is calculated and compared with the current judgment threshold to preliminarily determine whether the current soil moisture is in a response-required state and generate a preliminary judgment signal for local irrigation response; Collect soil moisture feedback data returned by the irrigation execution module and associate the feedback data with the preliminary judgment signal, the corresponding moisture change rate, the abnormal trend indicator data set and the judgment score to form a historical judgment sample data set; Based on the self-learning algorithm, the moisture change rate, abnormal trend indicator data, judgment score, judgment signal and soil moisture feedback data in the historical judgment sample data set are analyzed to train the model, dynamically adjust and optimize the current judgment threshold to adapt it to soil moisture and environmental changes, and generate the optimized current judgment threshold; Based on the optimized current judgment threshold, combined with the latest moisture change rate and abnormal trend indicators, the original soil moisture data is judged to generate the latest local irrigation response judgment signal. The latest judgment signal and soil moisture feedback data are continuously incorporated into the historical judgment sample data set, forming an adaptive optimization closed loop of judgment threshold and strategy. A response control module is used to automatically generate an irrigation equipment start instruction when the judgment signal indicates an abnormal state, otherwise continuously monitor and wait for a new judgment signal input; The irrigation execution module is used to execute irrigation operations according to the start-up instructions of the irrigation equipment, and at the same time collect soil moisture feedback data after the irrigation is completed, and send the soil moisture feedback data and soil moisture data synchronously to the feature extraction and analysis module for it to continuously optimize the judgment threshold and judgment strategy.

2. The smart agricultural environment monitoring system based on the Internet of Things according to claim 1 is characterized in that: Based on the original soil moisture data, combined with historical soil moisture data, historical ambient temperature data and preset recent period soil moisture feedback data, the original soil moisture data is statistically analyzed in sections to automatically extract the moisture change rate and abnormal trend indicators under multiple time windows, including: Based on the original soil moisture data and historical soil moisture data, the original soil moisture data is aggregated in segments based on a set time window to generate aggregated data for each target time period; Perform differential calculations on the aggregated data of each target time period and the historical soil moisture data to obtain the moisture change rate in each time window; Combined with historical ambient temperature data and preset recent soil moisture feedback data, a multidimensional cross-statistical analysis was conducted on the humidity change rate. The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index and humidity recovery lag index after irrigation were calculated to generate an abnormal trend indicator data set.

3. The smart agricultural environment monitoring system based on the Internet of Things according to claim 2 is characterized in that: Perform differential calculations on the aggregated data of each target time period and the historical soil moisture data to obtain the moisture change rate in each time window, including: Based on the aggregated data of each target time period, the current average soil moisture value of each target time period is extracted; According to each target time period, the historical average soil moisture value of the corresponding time period in the historical soil moisture data is extracted; The current average soil moisture value of each target time period is compared with its corresponding historical average soil moisture value. The relative change calculation method is used to divide the difference between the two by the historical average soil moisture value to obtain the humidity change rate in this time window.

4. The smart agricultural environment monitoring system based on the Internet of Things according to claim 3 is characterized in that: Combined with historical ambient temperature data and preset recent period soil moisture feedback data, a multi-dimensional cross-statistical analysis of humidity change rate was conducted. The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index, and humidity recovery lag index after irrigation were calculated to generate an abnormal trend indicator data set, including: The humidity change rate in each time window is combined with historical ambient temperature data and the correlation analysis method is used to obtain the temperature-humidity correlation index; The fluctuation amplitude and frequency of humidity change rate in each time window are statistically analyzed to generate humidity change variance index and abnormal fluctuation frequency index; Calculate the humidity change rate index and the duration index of the abnormality according to the humidity change rate data within the preset recent period; Conduct time series analysis on preset near-period soil moisture feedback data to determine the lag degree of moisture recovery and generate a moisture recovery lag index after irrigation. The humidity change rate index, humidity change variance index, abnormal fluctuation frequency index, continuous abnormal duration index, temperature and humidity correlation index and humidity recovery lag index after irrigation are combined into one set to generate an abnormal trend index dataset.

5. The smart agricultural environment monitoring system based on the Internet of Things according to claim 1 is characterized in that: Based on the moisture change rate and abnormal trend indicator dataset, a judgment score is calculated and compared with the current judgment threshold to preliminarily determine whether the current soil moisture is in a response-required state. This generates a preliminary judgment signal for the local irrigation response, including: Normalize each indicator in the humidity change rate and abnormal trend indicator data set in each time window to make each indicator have a unified numerical range and eliminate the influence of different dimensions to obtain a normalized indicator data set; Assign weights to the indicators in the normalized indicator dataset based on preset or self-learned weight parameters. The weight assignment can be dynamically adjusted based on historical irrigation results, enabling the judgment model to adaptively adjust the importance of different time periods and different features. Multiply each normalized indicator value by its corresponding weight one by one, and sum all weighted results to obtain the judgment score of each time window to form a judgment score sequence; In the judgment scoring sequence, the judgment score of each time window is compared with the current judgment threshold one by one. If the judgment score reaches or exceeds the current judgment threshold, a preliminary judgment signal of local irrigation that requires response is output, and the judgment score is recorded together with the relevant feature data for subsequent self-learning training.

6. The smart agricultural environment monitoring system based on the Internet of Things according to claim 5 is characterized in that: Based on the self-learning algorithm, the humidity change rate, abnormal trend indicator data, judgment score, judgment signal and soil moisture feedback data in the historical judgment sample data set are analyzed to train the model, dynamically adjust and optimize the current judgment threshold to adapt it to soil moisture and environmental changes, and generate the optimized current judgment threshold, including: For each group of samples in the historical judgment sample data set, the corresponding humidity change rate, abnormal trend index data, judgment score, preliminary judgment signal and soil moisture feedback data are extracted; Based on the relationship between the moisture change rate, abnormal trend index data, judgment score and soil moisture feedback data of each group of samples, the actual error of each judgment is calculated, and samples with large actual errors are given higher weights to form weighted statistical results; According to the weighted statistical results, a self-learning algorithm is used to analyze the setting deviation of the judgment threshold, and combined with the current soil moisture raw data, the current judgment threshold is adjusted to obtain the optimized current judgment threshold.

7. The smart agricultural environment monitoring system based on the Internet of Things according to claim 4 is characterized in that: The humidity change rate in each time window is combined with historical ambient temperature data and a correlation analysis method is used to obtain temperature and humidity correlation indicators, including: According to the humidity change rate in each time window, the historical ambient temperature data corresponding to the time window is extracted to construct the humidity change rate time series and the ambient temperature series; Normalize the humidity change rate time series and the ambient temperature series so that the two series have the same numerical range and eliminate the influence of different dimensions; The normalized humidity change rate time series and the ambient temperature series are paired one by one according to the time window order, the product sum of each paired point is calculated, and the respective mean and variance are statistically calculated to obtain the mean variance data; According to the mean-variance data, the Pearson correlation coefficient calculation method was used to solve the correlation coefficient of the two normalized series to quantify the linear correlation between the humidity change rate and the ambient temperature; The correlation coefficient is output as the temperature and humidity correlation index.

8. The smart agricultural environment monitoring system based on the Internet of Things according to claim 4 is characterized in that: Based on the humidity change rate data within the preset recent period, calculate the humidity change rate index and the duration of the abnormality, including: According to the humidity change rate data within the preset recent period, the humidity change rate time series is sorted in chronological order; For each continuous time interval in the humidity change rate time series, calculate the numerical change of the humidity change rate within the interval, and take the length of each time interval as the benchmark, divide the change by the time interval length to obtain the change rate of each period; The change rate of all consecutive time periods is averaged to obtain the humidity change rate index of the monitoring point in the recent period; In the humidity change rate time series, the normal interval threshold of the humidity change rate is set to determine whether each time point is abnormal. The time period continuously in the abnormal interval is accumulated as the abnormal duration to obtain the continuous abnormal duration index.

9. The smart agricultural environment monitoring system based on the Internet of Things according to claim 4 is characterized in that: Perform time series analysis on the preset recent soil moisture feedback data to determine the lag degree of moisture recovery and generate moisture recovery lag indicators after irrigation, including: According to the preset near-period soil moisture feedback data, it is sorted into a humidity feedback time series according to the collection time sequence; According to the humidity feedback time series, mark the humidity feedback value corresponding to the end time of the irrigation operation and each subsequent time; Set the interval threshold of the normal recovery interval of soil moisture, judge the recovery of the moisture feedback value based on the interval threshold, and identify the continuous time period from the lowest moisture point at the end of irrigation to the first time it reaches the normal recovery interval; The length of each continuous time period is taken as the humidity recovery hysteresis value and output as the humidity recovery hysteresis index after irrigation.

Images