Soil moisture detection method, device and electronic equipment

Abstract

The embodiment of the application relates to the technical field of soil moisture content detection, and provides a regional moisture content detection method, device and electronic equipment, the method comprising: acquiring current remote sensing satellite images and moisture content sensor data in a target region; inputting the current remote sensing satellite images into the latest target region moisture content inversion model to invert soil moisture content data of each pixel point in the target region; correcting the soil moisture content data of each pixel point based on the moisture content sensor data to obtain a moisture content detection result of the target region. Thus, the ground moisture content sensor, the remote sensing image receiver and various algorithms are combined, the advantages of the two measurement methods are combined, the measurement result is more accurate, the influence of various factors such as different regional topography, soil properties, vegetation coverage and vegetation types on the result is avoided, and the representativeness and accuracy of the detection result are improved.

Description

Technical Field

[0001] This invention relates to the field of soil moisture detection technology, and in particular to a regional moisture detection method, apparatus, and electronic equipment. Background Technology

[0002] Soil moisture refers to the water content in the topsoil layer used for crop cultivation. Soil moisture refers to the soil's humidity. Soil moisture content is the degree of dryness or wetness of the soil, i.e., the actual water content of the soil, which can be expressed as a percentage of soil moisture content relative to its dry weight: Soil moisture content = (water weight / dry weight) × 100%. It can also be expressed as a percentage of soil moisture content relative to field capacity or relative to saturation water, etc. Soil moisture affects field climate, soil aeration, and nutrient decomposition, and is one of the important conditions for soil microbial activity and crop growth and development. Soil moisture is affected by atmospheric conditions, soil type, vegetation, etc. The relative humidity of the soil indicates its water content and how much water it can retain, which is valuable for irrigation.

[0003] Currently, the main methods for measuring soil moisture include soil sample analysis, soil moisture sensor measurement, and satellite remote sensing technology. Soil sample analysis is the traditional method for measuring soil moisture. It mainly involves collecting soil samples, drying and weighing them in the laboratory, and then calculating the actual soil moisture content. The specific procedure includes: using a soil drill to collect a certain amount of soil sample at a selected location and depth. To prevent evaporation, the collected soil sample should be immediately placed in an aluminum box and labeled. It is then brought back to the laboratory for weighing, drying, re-weighing, and analysis to determine the soil moisture content. However, soil sample analysis has drawbacks such as: it is labor-intensive and time-consuming. Soil sample analysis requires collecting soil samples and performing a series of operations in the laboratory, including drying and weighing, which is cumbersome and time-consuming. This method is particularly inconvenient for situations requiring frequent or large-scale monitoring. Furthermore, it is difficult to obtain field data. Because soil samples need to be brought back to the laboratory for analysis, data cannot be obtained immediately on-site. This can cause delays in agricultural production that requires real-time monitoring of soil moisture changes. Limited representativeness: Soil sample analysis typically only targets specific sampling points, making it difficult to comprehensively reflect the soil moisture conditions of the entire field. Furthermore, the selection of sampling points and sampling methods can also affect the accuracy of the results. Human error: During sampling, transportation, processing, and analysis, human factors can lead to errors. For example, unclean sampling tools, uneven sample mixing, and inaccurate drying temperature control can all affect the accuracy of the final results.

[0004] Soil moisture sensors utilize electronic devices, such as soil moisture thermometers and wireless moisture monitoring systems, to monitor soil moisture in real time. Sensors can measure parameters such as the soil's dielectric constant, temperature, and conductivity, thereby indirectly calculating the soil's moisture content. However, soil moisture sensor measurement has several drawbacks and problems, including: high equipment cost (high-quality sensors are typically expensive, and multiple sensors are needed for comprehensive coverage, which is financially prohibitive for small-scale farmers or areas with limited resources); complex maintenance and management (sensors require regular maintenance and calibration to ensure accuracy, requiring specialized knowledge and skills that may be difficult for non-professionals); significant susceptibility to environmental factors (sensors may malfunction or produce errors under harsh conditions such as extreme temperatures, humidity, and electromagnetic interference); and limited representativeness (sensor measurements are limited to the deployed points and cannot comprehensively reflect the soil moisture conditions of the entire field). Furthermore, the selection and deployment methods of the monitoring points can also affect the accuracy of the results. Data interpretation requires specialized knowledge; sensor data typically needs to be comprehensively analyzed in conjunction with multiple factors such as weather and crop growth conditions to arrive at effective irrigation recommendations. This can be challenging for ordinary farmers.

[0005] Satellite remote sensing technology utilizes sensors mounted on high-altitude platforms such as satellites and drones to conduct long-distance, non-contact observations of farmland. By capturing spectral and textural features of the farmland surface, it enables the monitoring and assessment of soil moisture. This method has advantages such as wide monitoring range, rapid data acquisition, and strong analytical processing capabilities. Remote sensing inversion of soil moisture is mainly achieved through spectral characteristic analysis, vegetation index analysis, and microwave remote sensing. Spectral reflectance characteristics: Soil moisture content affects its spectral reflectance characteristics, especially in the near-infrared and short-wave infrared bands, which are highly sensitive to changes in soil moisture. By measuring the reflectance of soil in different bands, the soil moisture content can be indirectly inferred. Vegetation index: The growth status of vegetation is closely related to soil moisture conditions. When soil moisture is sufficient, vegetation grows vigorously, resulting in a higher vegetation index; conversely, it decreases. By analyzing the changing trends of the vegetation index, soil moisture can be indirectly assessed. Microwave remote sensing: This method uses changes in the soil dielectric constant to monitor soil moisture. Soil moisture exhibits strong absorption and scattering of microwave signals, allowing for the inference of soil moisture content by measuring the propagation characteristics of microwave signals in the soil. However, remote sensing technology has drawbacks and limitations, such as: limited data resolution and accuracy. While remote sensing can cover large areas, its data resolution and accuracy may be limited by factors such as satellite orbit and sensor performance. This may not meet the needs of small-area or fine-grained soil moisture monitoring. Weather conditions, such as cloud cover and rainfall, can prevent the acquisition of remote sensing data or degrade its quality, affecting the accuracy of monitoring results. Data processing is complex: remote sensing data is large in volume and diverse in type, requiring highly specialized skills and knowledge for effective processing and analysis. This may pose a barrier for ordinary users. Summary of the Invention

[0006] This invention provides a regional soil moisture detection method, device, and electronic equipment to address the shortcomings of existing technologies, such as high cost, low detection accuracy, significant environmental influence, and low representativeness of detection results, thereby achieving low-cost, high-precision, low-impact, and highly representative soil moisture measurement.

[0007] This invention provides a method for detecting regional soil moisture, comprising:

[0008] Acquire current remote sensing satellite imagery and soil moisture sensor data within the target area;

[0009] The current remote sensing satellite image is input into the latest target area soil moisture inversion model so that the latest target area soil moisture inversion model can invert the soil moisture data of each pixel in the target area;

[0010] Based on the soil moisture sensor data, the soil moisture data of each pixel is corrected to obtain the soil moisture detection result of the target area;

[0011] The soil moisture inversion model for the target area is trained based on the following steps:

[0012] Acquire soil moisture measurement data in the sample soil and the first remote sensing satellite imagery for the corresponding measurement time period;

[0013] Band calculations are performed on the first remote sensing satellite image to obtain multiple band calculation results reflected in the first remote sensing satellite image, wherein each band calculation result includes multiple parameter indices;

[0014] Randomly extract an inversion dataset and a test dataset from the soil moisture measurement data and the multiple band calculation results, respectively. The inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model.

[0015] The inversion dataset and the results of the multiple band operations are input into the full subset filtering algorithm to filter and obtain the optimal combination of variables.

[0016] The optimal combination of variables and the inversion dataset are input into a variety of preset algorithm models to obtain the inversion model established by each algorithm.

[0017] The test dataset is input into the inversion model established by each algorithm, and the inversion models established by the various algorithms are tested by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

[0018] In one possible implementation, the method further includes:

[0019] The ratio of the soil moisture data of each pixel to the corresponding moisture sensor data is used as a correction coefficient;

[0020] The ratio of the soil moisture data of each pixel to the correction coefficient is used as the soil moisture detection result of the target area.

[0021] In one possible implementation, the method further includes:

[0022] Based on preset rules, the number and location of soil sampling plots are determined within the target area, resulting in multiple soil sampling plots.

[0023] Soil moisture measurement data were acquired using soil moisture sensors in the multiple soil sampling plots.

[0024] In one possible implementation, the method further includes:

[0025] The reflectance in the blue band, green band, red band, and near-infrared band displayed in the first remote sensing satellite image is obtained.

[0026] The calculation results of multiple bands reflected in the first remote sensing satellite image are calculated based on the blue light band reflectance, green light band reflectance, red light band reflectance and near-infrared band reflectance and the soil line coefficient of the sample soil.

[0027] In one possible implementation, the method further includes:

[0028] Based on a preset ratio and a preset number of extractions, inversion datasets and test datasets are randomly extracted from the soil moisture measurement data and the calculation results of the multiple bands, respectively, to obtain the inversion dataset and test dataset for each extraction.

[0029] In one possible implementation, the method further includes:

[0030] The model is trained based on the inversion dataset and test dataset extracted each time;

[0031] The goodness-of-fit of the model trained based on each extracted inversion dataset and test dataset is evaluated, and the optimal combination of variables with the best goodness-of-fit is selected.

[0032] In one possible implementation, the method further includes:

[0033] The optimal combination of variables and the inversion dataset extracted each time are input into a variety of preset algorithm models to obtain multiple inversion models established by each algorithm.

[0034] In one possible implementation, the method further includes:

[0035] The test dataset is input into multiple inversion models established by each algorithm, and the inversion models are verified based on goodness of fit and root mean square error.

[0036] The inversion model with the highest goodness of fit and the smallest root mean square error is selected as the optimal soil moisture inversion model for the target area.

[0037] Alternatively, the relative error coefficient of each inversion model can be calculated based on the goodness of fit and root mean square error, and the inversion model with the highest relative error coefficient can be selected as the optimal soil moisture inversion model for the target area.

[0038] The present invention also provides a regional soil moisture detection device, comprising the following modules:

[0039] The acquisition module is used to acquire current remote sensing satellite imagery and soil moisture sensor data within the target area;

[0040] The detection module is used to input the current remote sensing satellite image into the latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model can invert the soil moisture data of each pixel in the target area;

[0041] The correction module is used to correct the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area;

[0042] The model training module is used to acquire soil moisture measurement data and first remote sensing satellite images within the corresponding measurement time period; perform band calculations on the first remote sensing satellite images to obtain multiple band calculation results reflected in the first remote sensing satellite images, wherein each band calculation result includes multiple parameter indices; randomly extract an inversion dataset and a test dataset from the soil moisture measurement data and the multiple band calculation results, wherein the inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model; input the inversion dataset and the multiple band calculation results into a full subset filtering algorithm to filter out the optimal variable combination; input the optimal variable combination and the inversion dataset into a preset set of multiple algorithm models to obtain an inversion model established by each algorithm; input the test dataset into the inversion models established by each algorithm, and test the inversion models established by the multiple algorithms by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

[0043] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the regional soil moisture detection method as described above.

[0044] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the regional soil moisture detection method as described above.

[0045] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the regional soil moisture detection method as described above.

[0046] The present invention provides a regional soil moisture detection method, apparatus, and electronic device, which acquires current remote sensing satellite imagery and soil moisture sensor data within a target area; inputs the current remote sensing satellite imagery into the latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model inverts the soil moisture data of each pixel within the target area; and corrects the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area. The target area soil moisture inversion model is trained based on the following steps: acquiring soil moisture measurement data from a sample soil and a first remote sensing satellite imagery within the corresponding measurement time period; performing band calculations on the first remote sensing satellite imagery to obtain the multiple band calculation results reflected in the first remote sensing satellite imagery. Each band calculation result includes multiple parameter indices; an inversion dataset and a test dataset are randomly extracted from the soil moisture measurement data and the multiple band calculation results, respectively. The inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model; the inversion dataset and the multiple band calculation results are input into a full subset selection algorithm to select the optimal variable combination; the optimal variable combination and the inversion dataset are input into a variety of preset algorithm models to obtain the inversion model established by each algorithm; the test dataset is input into the inversion model established by each algorithm, and the inversion models established by the various algorithms are tested by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area. Compared to existing technologies that suffer from high cost, low detection accuracy, susceptibility to environmental influences, and low representativeness of results, this solution combines the advantages of both methods and mitigates their disadvantages by integrating a ground-based soil moisture sensor, a remote sensing image receiver, and multiple algorithms. It allows for convenient remodeling of each measurement area, resulting in more accurate measurements and avoiding the influence of factors such as topography, soil properties, vegetation cover, and vegetation type. Utilizing remote sensing satellites for quantitative soil moisture inversion across the entire area improves the representativeness and accuracy of the results. Furthermore, the soil moisture sensor compensates for the time gap between acquiring two satellite images, enabling dynamic monitoring of soil moisture and achieving low-cost, high-precision, low-impact, and highly representative soil moisture measurements. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0048] Figure 1This is a flowchart illustrating the regional soil moisture detection method provided by the present invention.

[0049] Figure 2 This is a flowchart illustrating the training method for the target area soil moisture inversion model provided by the present invention.

[0050] Figure 3 This is a schematic diagram illustrating the instrument deployment principle and usage process provided by the present invention.

[0051] Figure 4 This is a schematic diagram of the regional soil moisture detection output results provided by the present invention.

[0052] Figure 5 This is a schematic diagram of the structure of the regional soil moisture detection device provided by the present invention.

[0053] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0055] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0056] Figure 1 This is a flowchart illustrating the regional soil moisture detection method provided by the present invention, as shown below. Figure 1 As shown, the method specifically includes:

[0057] S11. Acquire current remote sensing satellite imagery and soil moisture sensor data within the target area.

[0058] In this embodiment of the invention, a target region soil moisture inversion model is first trained for the target region. The specific training method is as follows: Figure 2 The corresponding embodiments are described in detail. After the soil moisture inversion model for the target area is trained, the current remote sensing satellite imagery and soil moisture sensor data of the target area to be detected are acquired.

[0059] S12. Input the current remote sensing satellite image into the latest target area soil moisture inversion model so that the latest target area soil moisture inversion model can invert the soil moisture data of each pixel in the target area.

[0060] During model training, the system automatically records the best inversion model (the latest target area soil moisture inversion model) for each sampling time. After recording the best inversion model, the data from the next measurement is input into the latest target area soil moisture inversion model.

[0061] The current remote sensing satellite imagery is input into the latest target area soil moisture inversion model so that the latest target area soil moisture inversion model can invert the soil moisture data of each pixel in the target area.

[0062] S13. Based on the soil moisture sensor data, the soil moisture data of each pixel is corrected to obtain the soil moisture detection result of the target area.

[0063] The ratio of soil moisture data at each pixel to the corresponding soil moisture sensor data is used as the correction coefficient; the ratio of soil moisture data at each pixel to the correction coefficient is used as the soil moisture detection result of the target area.

[0064] The calibration steps are as follows:

[0065] 1) The instrument records the soil moisture values ​​at different depths of the set points every 30 minutes.

[0066] 2) Using the latest remote sensing images combined with the best inversion model, the soil moisture results at the time the images were acquired were inverted.

[0067] 3) Obtain the latest remote sensing image time, and use the soil moisture value with the closest recording time as the correction value. If the times are the same, take the average value of the two times as the correction value.

[0068] 4) Obtain the soil moisture inversion results at the instrument location and obtain the correction coefficient. Correction coefficient = soil moisture inversion results at the instrument location / correction value.

[0069] 5) Dividing the data of each pixel in the entire inversion result by the correction coefficient yields the corrected true value of all pixels, thus obtaining the corrected soil moisture sensor data for the measured area. The final regional soil moisture detection output is as follows: Figure 4 As shown.

[0070] Optionally, if the current remote sensing satellite image of the target area is not acquired within a preset time period (e.g., 30 minutes), the previously obtained soil moisture data is corrected based on the soil moisture sensor data.

[0071] Specifically, during the time interval between the acquisition of two images, the instrument can record soil moisture values ​​at different depths of the set points every 30 minutes to correct the soil moisture results of the entire area at that time, thereby obtaining regional soil moisture data every 30 minutes.

[0072] The calibration steps are as follows:

[0073] 1) The instrument records the soil moisture values ​​at different depths of the set points every 30 minutes.

[0074] 2) Use the soil moisture value obtained each time as the correction value.

[0075] 3) Obtain the soil moisture inversion results from the latest remote sensing image of the instrument location, and obtain the correction coefficient. The correction coefficient = soil moisture inversion results from the latest remote sensing image of the instrument location / correction value.

[0076] 4) Divide the data of each pixel in the soil moisture inversion results obtained from all the latest remote sensing images by the correction coefficient to obtain the corrected true value of all pixels, thereby obtaining the soil moisture result value of the measurement area for that period.

[0077] This invention combines ground-based soil moisture sensors and satellite remote sensing technology to determine the optimal regional soil moisture retrieval model. Utilizing the real-time performance, adaptability, and data continuity of ground-based sensors, it effectively compensates for the data gaps between two satellite remote sensing images, effectively improving the resolution and accuracy of the results. Utilizing satellite remote sensing technology, with its wide monitoring range and strong comprehensive analysis capabilities, it effectively enhances the representativeness of the results, reduces the impact of equipment deployment environment on the results, reduces the number of devices deployed, thereby lowering equipment costs and minimizing the impact of deploying multiple devices on farmland.

[0078] The regional soil moisture detection method provided by this invention acquires current remote sensing satellite imagery and soil moisture sensor data within a target area; inputs the current remote sensing satellite imagery into the latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model inverts the soil moisture data of each pixel within the target area; and corrects the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area. Compared to existing technologies that suffer from high cost, low detection accuracy, susceptibility to environmental influences, and low representativeness of results, this method combines the advantages of both methods and compensates for their shortcomings by integrating a ground soil moisture sensor, a remote sensing image receiver, and multiple algorithms. It allows for convenient remodeling of each measurement area, resulting in more accurate measurements and avoiding the influence of factors such as topography, soil properties, vegetation cover, and vegetation type. Utilizing remote sensing satellites for quantitative soil moisture inversion across the entire area improves the representativeness and accuracy of the results. Furthermore, the soil moisture sensor compensates for the time interval between two satellite image acquisitions, enabling dynamic monitoring of soil moisture and achieving low-cost, high-precision, low-impact, and highly representative soil moisture measurement.

[0079] Figure 2 This is a flowchart illustrating the training method for the target area soil moisture inversion model provided by the present invention, as shown below. Figure 2 As shown, the method specifically includes:

[0080] S21. Obtain soil moisture measurement data in the sample soil and the first remote sensing satellite image of the corresponding measurement time period.

[0081] The embodiments of the present invention are combined with Figure 3 The schematic diagram illustrating the instrument's deployment principle and usage process is provided below. The system in this embodiment is integrated into a soil moisture measurement instrument (electronic device), whose hardware also includes a soil moisture sensor. First, based on preset rules, the number and location of soil sampling plots are determined within the target area, resulting in multiple soil sampling plots. Soil moisture measurement data are then acquired within each of these multiple sampling plots using the soil moisture sensor.

[0082] Specifically, the number of quadrats should be determined within the sample measurement area. The requirement is at least one quadrat for every 60 mu (approximately 4 hectares). For example, if the area to be measured is 300 mu (approximately 20 hectares), then 300 mu / 60 mu = 5 quadrats should be selected. If the area to be measured is 320 mu (approximately 21 hectares), then 6 quadrats should be selected.

[0083] Furthermore, determine the locations of quadrats within the measurement area. Each quadrat is a 3m x 3m square, and its location must be at least 30m from the boundary of the measurement area. Quadrats should be evenly and randomly distributed within the measurement area. Additional quadrats may be added as needed for areas with special terrain or land properties within the measurement area. For example, if there are areas with significantly higher or lower elevations, or areas with high salinity, additional quadrats may be added in these special areas.

[0084] Furthermore, five sampling points can be randomly selected within each quadrat. The instrument's ground moisture sensor is used to measure soil moisture within the quadrat. Before measurement, quadrat 1 through quadrat N are created in the instrument. When measuring a quadrat, the corresponding quadrat number is opened. The measurement method involves inserting the instrument's ground moisture sensor into the soil at the sampling point to a depth equal to the instrument's standard scale, 80cm, for at least 30 seconds. This will obtain soil moisture data at depths of 20cm, 40cm, 60cm, and 80cm, as well as the coordinates of the sampling point. The instrument will automatically record these data as points 1 through N within quadrat N. For example, when measuring the first point in quadrat 1, quadrat 1 is created. After the first measurement, the instrument automatically records the data as point 1 in quadrat 1. When measuring the second point in quadrat 1, after the second measurement, the instrument automatically records the data as point 2 in quadrat 1, and so on.

[0085] Furthermore, the measurement of quadrats within the same measurement area should be completed within 2 days as much as possible. After all quadrats are measured, the instrument will use an algorithm to screen all data and identify abnormal data. Users can confirm or delete abnormal data according to the actual situation.

[0086] S22. Perform band calculations on the first remote sensing satellite image to obtain the results of multiple band calculations reflected in the first remote sensing satellite image.

[0087] The reflectance in the blue, green, red, and near-infrared bands of the first remote sensing satellite imagery is acquired. Based on these reflectances and the soil line coefficient of the sample soil, the calculation results for multiple bands in the first remote sensing satellite imagery are obtained. Each band calculation result includes multiple parameter indices.

[0088] Specifically, after the user confirms the data, the instrument will acquire remote sensing images for the measurement period (at least four bands: red, green, blue, and near-infrared, with a spatial resolution ≤10m). The acquired remote sensing images will then undergo the following band calculations:

[0089] In the following formula and These correspond to the blue, green, red, and near-infrared reflectance of the remote sensing image, respectively. These parameters are inherent to the remote sensing image and can be read directly. a and b represent soil linear coefficients.

[0090] Normalized Difference Vegetation Index (NDVI)

[0091]

[0092] Enhanced Vegetation Index (EVI)

[0093]

[0094] Ratio Vegetation Index (RVI)

[0095]

[0096] Difference Vegetation Index (DVI)

[0097]

[0098] Soil-modified vegetation index (SAVI)

[0099]

[0100] Soil Improvement and Adjustment of Vegetation Index (MSAVI)

[0101]

[0102] Optimize the soil-regulated vegetation index (OSAVI)

[0103]

[0104] Improved Simple Ratio Vegetation Index (MSR)

[0105]

[0106] Structure-Insensitive Pigment Index (SIPI)

[0107]

[0108] Modified Nonlinear Vegetation Index (MNVI)

[0109]

[0110] Triangular Vegetation Index (TVI)

[0111]

[0112] Green chlorophyll index (GLI)

[0113]

[0114] Modified chlorophyll absorbance reflectance index 2 (MCARI2)

[0115]

[0116] Salinity Index (SI-T)

[0117]

[0118] Salinity index (S3)

[0119]

[0120] Salinity index (S5)

[0121]

[0122] Salinity index (S6)

[0123]

[0124] Soil Adjustment Index (TSAVI)

[0125]

[0126] Building Index (BI)

[0127]

[0128] Bare Soil Index (SI2)

[0129]

[0130] Vertical drought index (PDI)

[0131]

[0132] Soil-Adjusted Vegetation Index (SAVI2)

[0133]

[0134] S23. Randomly extract the inversion dataset and the test dataset from the soil moisture measurement data and the multiple band calculation results, respectively.

[0135] S24. Input the inversion dataset and the results of the multiple band operations into the full subset filtering algorithm to filter and obtain the optimal variable combination.

[0136] Based on a preset ratio and a preset number of extractions, inversion datasets and test datasets are randomly extracted from soil moisture measurement data and multiple band calculation results, respectively, to obtain the inversion dataset and test dataset for each extraction. The inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model. The model is trained based on each extracted inversion dataset and test dataset; the goodness-of-fit of the model trained based on each extracted inversion dataset and test dataset is evaluated, and the optimal variable combination with the best goodness-of-fit is selected.

[0137] Specifically, after processing the acquired image bands, 80% of the sampled data can be randomly selected as inversion data, and the remaining 20% ​​as test data. This random selection is repeated 100 times. The method for selecting data once is as follows: randomly select 80% from all sampled data as inversion data, and the remaining 20% ​​after this selection is used as test data. The inversion data is then used to input the aforementioned multiple band calculation formulas to calculate the corresponding results, which are then input into the full subset selection model. After inputting the extracted data into the full subset selection model, many model combinations will appear. The test data is used to perform a coefficient of determination test on all model combinations. The combination with the highest coefficient of determination is the best model selection after this data extraction, and this combination and its coefficient of determination must be recorded.

[0138] The above is the result of one optimal variable selection. To obtain the optimal variable selection results, we need to perform 100 optimal variable selections, which means randomly selecting 100 times to obtain 100 optimal variables and their corresponding coefficients of determination. The optimal variable selection result with the highest coefficient of determination is the selected optimal variable combination.

[0139] The core of the All-Subsets Selection algorithm lies in iterating through all possible combinations of feature variables, evaluating the model for each combination, and ultimately selecting the optimal feature subset. This method aims to find the best-performing model on the current dataset by comprehensively considering all possible feature combinations. This comprehensiveness gives the All-Subsets Selection algorithm high flexibility and accuracy in feature selection. Its steps are as follows:

[0140] 1) Generating all subsets: Assuming there are p feature variables, the full subset selection algorithm will generate all possible feature subsets. This includes univariate subsets containing only one feature, bivariate subsets containing two features, and so on up to the entire set containing all p features. The total number of subsets is 2^p-1 (including the entire set, excluding the empty set).

[0141] 2) Model Fitting: For each generated feature subset, the same modeling method (such as linear regression, logistic regression, decision tree, etc.) is used to fit the model. In this embodiment of the invention, the methods used are linear regression and random forest. This step needs to be performed separately for each subset, so the computational cost increases rapidly with the number of features.

[0142] 3) Performance Evaluation: The coefficient of determination (R²) is used to evaluate the performance of each model. These metrics are used to measure the model's goodness of fit, complexity, and predictive ability. Based on the evaluation results, different feature subsets are compared and ranked to identify the best-performing feature subset. The coefficient of determination (R²), also known as goodness of fit, is an indicator that measures a model's ability to predict the target variable. It represents the degree to which the independent variables in the model explain the dependent variable, that is, the percentage of variation in the dependent variable that the independent variables in the model can explain. The formula for calculating R² is: R² = 1 - (SSE / SST).

[0143] Here, SSE stands for Sum of Squared Errors, representing the sum of squares of the differences between the model's predicted values ​​and the actual values; SST stands for Total Sum of Squares, representing the sum of squares of the differences between the actual values ​​and the mean. The value of R² is between 0 and 1, with a value closer to 1 indicating a higher goodness of fit, meaning that the independent variables in the model can better explain the variation in the dependent variable.

[0144] 4) Selecting the optimal subset: Based on the performance evaluation results, the instrument selects the optimal feature subset. This subset is usually the one that performs best under the current evaluation criteria, that is, it best balances the model's goodness of fit and complexity.

[0145] S25. Input the optimal variable combination and the inversion dataset into a preset set of multiple algorithm models to obtain the inversion model established by each algorithm.

[0146] The optimal combination of variables and the inversion dataset extracted each time are input into multiple preset algorithm models to obtain multiple inversion models built by each algorithm.

[0147] Specifically, the optimal combination of variables selected and 100 randomly selected inversion data points are input into the instrument's partial least squares regression algorithm (a common algorithm), backpropagation neural network model (a common algorithm), genetic algorithm-optimized backpropagation neural network model (a common algorithm), and random forest algorithm (a common algorithm), resulting in 100 models built by each algorithm. Here, 100 data points can be re-sampled from all the data according to a preset ratio; for example, 80% for inversion data and 20% for test data.

[0148] Partial Least Squares Regression (PLSR) is a multivariate statistical data analysis method that establishes a linear relationship model between independent and dependent variables. PLSR projects the original data into a new space, extracting the composite variable (i.e., component) that best explains the dependent variable, and then performs regression modeling. This method is particularly suitable for datasets with many variables and multicollinearity. Its advantages include effectively handling multicollinearity, extracting important information, and simplifying the model. Its disadvantage is that the model's interpretability may be weak because the components may be difficult to interpret directly.

[0149] Backpropagation Neural Network (BPNN) is a multi-layer feedforward neural network trained using the backpropagation algorithm. This algorithm utilizes the chain rule to calculate the gradient of the loss function with respect to each weight and updates the weights to minimize the loss function. BPNN can learn complex nonlinear relationships between inputs and outputs. Its advantages include powerful nonlinear modeling capabilities and the ability to handle complex problems. Its disadvantages include potentially slow training and a susceptibility to overfitting.

[0150] Genetic Algorithm for Optimizing Backpropagation Neural Network (BPNN) is an optimization algorithm that simulates the biological evolutionary process. It searches for the optimal solution through operations such as selection, crossover, and mutation. Applying genetic algorithms to BPNN optimization can improve the weight initialization of BPNNs, thereby increasing training efficiency and model performance. Its advantage lies in combining the global search capability of genetic algorithms with the local search capability of BPNNs to enhance model performance. Its disadvantages include high computational complexity and the need for appropriate genetic algorithm parameter settings.

[0151] Random Forest is an ensemble learning method that constructs multiple decision trees and averages or majority votes their predictions. When constructing each tree, Random Forest increases tree diversity by introducing randomness into the samples and features. Its advantages include high prediction accuracy, the ability to handle high-dimensional data, and good robustness to outliers and noise. Its disadvantages include poor model interpretability and potentially slow training processes.

[0152] S26. Input the test dataset into the inversion model established by each algorithm, and test the inversion models established by the multiple algorithms by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

[0153] The test dataset was input into multiple inversion models established by each algorithm, and the inversion models were verified based on the goodness of fit and root mean square error. The inversion model with the highest goodness of fit and the smallest root mean square error was selected as the best soil moisture inversion model for the target area.

[0154] Specifically, the remaining 20% ​​of the test data from 100 samplings is input into 100 models built for each algorithm. The inversion models are tested by calculating the coefficient of determination (R²) and root mean square error (MSE). If the optimal inversion model selected by the two parameters is the same, then this model is the optimal inversion model for the measurement area. If the optimal inversion models selected by the two parameters are different, the difference between the two coefficients is calculated, and the model with the larger difference is the optimal inversion model.

[0155] The relative error coefficient (REC) of each model is calculated using the coefficient of determination (R²) and root mean square error (MSE). The model with the highest relative error coefficient (REC) is the best inversion model.

[0156] The coefficient of determination, also known as the goodness of fit, is an indicator that measures a model's ability to predict the target variable. It represents the extent to which the independent variables in the model explain the variation in the dependent variable, that is, the percentage of variation in the dependent variable that the independent variables in the model can explain. The formula for calculating R² is: R² = 1 - (SSE / SST).

[0157] Here, SSE stands for Sum of Squared Errors, representing the sum of squares of the differences between the model's predicted values ​​and the actual values; SST stands for Total Sum of Squares, representing the sum of squares of the differences between the actual values ​​and the mean. The value of R² is between 0 and 1, with a value closer to 1 indicating a higher goodness of fit, meaning that the independent variables in the model can better explain the variation in the dependent variable.

[0158] Root mean square error (MSE) is a metric for measuring the accuracy of a model's predictions. It is represented by the square root of the average of the sum of squares of the differences between the model's predicted and actual values. The formula for calculating MSE is: .

[0159] Where SSE is the sum of squared residuals and n is the number of samples. The smaller the MSE value, the higher the prediction accuracy of the model.

[0160] Alternatively, the relative error coefficient of each inversion model can be calculated based on the goodness of fit and root mean square error, and the inversion model with the highest relative error coefficient can be taken as the best soil moisture inversion model for the target area.

[0161] The relative error coefficient (REC) is a coefficient obtained by normalizing the coefficient of determination (R²) and the root mean square error (RMSE) according to their reference significance, and is used to judge the model fitting accuracy.

[0162] The formula for calculating REC is: REC = .

[0163] in, For a model result, For all models The minimum value in the results For all models The maximum value in the result. For a model result, For all models The minimum value in the results For all models The maximum value in the result.

[0164] Furthermore, the instrument automatically records the optimal inversion model for each sampling time. After recording the optimal inversion model, the data from the next measurement is input into the latest optimal inversion model. Multiple tasks can be performed at different times in the same measurement area, and the instrument automatically records the optimal inversion model multiple times according to the time.

[0165] The instrument is deployed within the measurement area, and each time it acquires the latest remote sensing satellite image, the image is input into the instrument's stored optimal inversion model with the closest time frame. The soil moisture value for each pixel in the entire measurement area is then retrieved, and the moisture sensor data from the image acquisition time is used to correct the moisture results for the measurement area.

[0166] The calibration steps are as follows:

[0167] 1) The instrument records the soil moisture values ​​at different depths of the set points every 30 minutes.

[0168] 2) Using the latest remote sensing images combined with the best inversion model, the soil moisture results at the time the images were acquired were inverted.

[0169] 3) Obtain the latest remote sensing image time, and use the soil moisture value with the closest recording time as the correction value. If the times are the same, take the average value of the two times as the correction value.

[0170] 4) Obtain the soil moisture inversion results at the instrument location and obtain the correction coefficient. Correction coefficient = soil moisture inversion results at the instrument location / correction value.

[0171] 5) Dividing the data of each pixel in all inversion results by the correction coefficient yields the corrected true value of all pixels, thus obtaining the corrected soil moisture sensor data and the soil moisture result value of the measurement area.

[0172] Furthermore, during the time interval between the acquisition of the two images, the instrument will record the soil moisture values ​​at different depths of the deployed points every 30 minutes to correct the soil moisture results of the entire area at that time, thereby obtaining regional soil moisture data every 30 minutes.

[0173] The calibration steps are as follows:

[0174] 1) The instrument records the soil moisture values ​​at different depths of the set points every 30 minutes.

[0175] 2) Use the soil moisture value obtained each time as the correction value.

[0176] 3) Obtain the soil moisture inversion results from the latest remote sensing image of the instrument location, and obtain the correction coefficient. The correction coefficient = soil moisture inversion results from the latest remote sensing image of the instrument location / correction value.

[0177] 4) Divide the data of each pixel in the soil moisture inversion results obtained from all the latest remote sensing images by the correction coefficient to obtain the corrected true value of all pixels, thereby obtaining the soil moisture result value of the measurement area for that period.

[0178] The method for training a soil moisture inversion model in a target area provided in this invention involves acquiring soil moisture measurement data from a sample soil and a first remote sensing satellite image within the corresponding measurement time period; performing band calculations on the first remote sensing satellite image to obtain multiple band calculation results reflected in the first remote sensing satellite image, wherein each band calculation result includes multiple parameter indices; randomly extracting an inversion dataset and a test dataset from the soil moisture measurement data and the multiple band calculation results, wherein the inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model; inputting the inversion dataset and the multiple band calculation results into a full subset filtering algorithm to filter for the optimal variable combination; inputting the optimal variable combination and the inversion dataset into a preset set of multiple algorithm models to obtain an inversion model established by each algorithm; inputting the test dataset into the inversion models established by each algorithm, and testing the inversion models established by the multiple algorithms by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area. This method combines the advantages of both ground-based soil moisture sensors and remote sensing image receivers with multiple algorithms, thus mitigating the disadvantages of both methods. It allows for convenient remodeling of each measurement area, leading to more accurate instrument measurements and avoiding the influence of factors such as regional topography, soil properties, vegetation cover, and vegetation type. Different regions, soil properties, vegetation types and canopies, and topography result in different combinations of sensitive bands used for soil moisture inversion. Therefore, by first selecting points for on-site measurement and then filtering characteristic variables, the most suitable inversion input parameters for the specific region can be obtained. Since actual monitoring after instrument deployment may differ slightly from point sampling—for example, date, temperature, air pressure, and humidity can affect inversion accuracy—the addition of ground sensors can correct the remote sensing inversion results, making them closer to the current actual values. The use of remote sensing imagery fundamentally solves the limitations of ground sensor monitoring range and the problem of data gaps between two image sets. It can accurately and quantitatively retrieve the actual values ​​of each pixel in a remote sensing image.

[0179] The regional soil moisture detection device provided by the present invention will be described below. The regional soil moisture detection device described below can be referred to in correspondence with the regional soil moisture detection method described above.

[0180] Figure 5 This is a schematic diagram of the structure of the regional soil moisture detection device provided by the present invention, specifically including:

[0181] The acquisition module 501 is used to acquire current remote sensing satellite imagery and soil moisture sensor data within the target area. For detailed explanations, please refer to the relevant descriptions in the above method embodiments; they will not be repeated here.

[0182] The detection module 502 is used to input the current remote sensing satellite image into the latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model can invert the soil moisture data of each pixel in the target area. For detailed explanation, please refer to the relevant descriptions in the above method embodiments, which will not be repeated here.

[0183] The correction module 503 is used to correct the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area. For detailed explanation, please refer to the relevant descriptions in the above method embodiments, which will not be repeated here.

[0184] The model training module 504 is used to acquire soil moisture measurement data in the sample soil and the first remote sensing satellite image within the corresponding measurement time period; perform band calculations on the first remote sensing satellite image to obtain multiple band calculation results reflected in the first remote sensing satellite image, wherein each band calculation result includes multiple parameter indices; randomly extract an inversion dataset and a test dataset from the soil moisture measurement data and the multiple band calculation results, wherein the inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model; input the inversion dataset and the multiple band calculation results into a full subset filtering algorithm to filter out the optimal variable combination; input the optimal variable combination and the inversion dataset into a preset set of multiple algorithm models to obtain an inversion model established by each algorithm; input the test dataset into the inversion model established by each algorithm, and test the inversion models established by the multiple algorithms by calculating the coefficient of determination and root mean square error to determine the optimal target area soil moisture inversion model. For detailed explanations, please refer to the relevant descriptions corresponding to the above method embodiments, which will not be repeated here.

[0185] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6As shown, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communications bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other through the communications bus 840. The processor 810 can call logical instructions in the memory 830 to execute a regional soil moisture detection method. This method includes: acquiring current remote sensing satellite imagery and soil moisture sensor data within a target area; inputting the current remote sensing satellite imagery into the latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model inverts the soil moisture data of each pixel within the target area; correcting the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area; wherein the target area soil moisture inversion model is trained based on the following steps: acquiring soil moisture measurement data in a sample soil and a first remote sensing satellite imagery within the corresponding measurement time period; performing band calculations on the first remote sensing satellite imagery to obtain multiple bands represented in the first remote sensing satellite imagery. The calculation results of each band include multiple parameter indices. An inversion dataset and a test dataset are randomly extracted from the soil moisture measurement data and the multiple band calculation results, respectively. The inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model. The inversion dataset and the multiple band calculation results are input into a full subset selection algorithm to select the optimal variable combination. The optimal variable combination and the inversion dataset are input into multiple preset algorithm models to obtain the inversion model established by each algorithm. The test dataset is input into the inversion model established by each algorithm, and the inversion models established by the multiple algorithms are tested by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

[0186] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0187] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the regional soil moisture detection method provided by the above methods. The method includes: acquiring current remote sensing satellite imagery and soil moisture sensor data within a target area; inputting the current remote sensing satellite imagery into the latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model inverts the soil moisture data of each pixel within the target area; correcting the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area; wherein the target area soil moisture inversion model is trained based on the following steps: acquiring soil moisture measurement data in sample soil and a first remote sensing satellite imagery within the corresponding measurement time period; and inputting the current remote sensing satellite imagery into the latest target area soil moisture inversion model to invert the soil moisture data of each pixel within the target area. A remote sensing satellite image is used for band calculations to obtain multiple band calculation results reflected in the first remote sensing satellite image, wherein each band calculation result includes multiple parameter indices; an inversion dataset and a test dataset are randomly extracted from the soil moisture measurement data and the multiple band calculation results, respectively, wherein the inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model; the inversion dataset and the multiple band calculation results are input into a full subset selection algorithm to select the optimal variable combination; the optimal variable combination and the inversion dataset are input into a variety of preset algorithm models to obtain an inversion model established by each algorithm; the test dataset is input into the inversion model established by each algorithm respectively, and the inversion models established by the various algorithms are tested by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

[0188] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, this computer program performs the regional soil moisture detection method provided by the methods described above. This method includes: acquiring current remote sensing satellite imagery and soil moisture sensor data within a target area; inputting the current remote sensing satellite imagery into a latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model inverts the soil moisture data of each pixel within the target area; correcting the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area; wherein the target area soil moisture inversion model is trained based on the following steps: acquiring soil moisture measurement data in a sample soil and a first remote sensing satellite imagery within the corresponding measurement time period; performing band calculations on the first remote sensing satellite imagery; The calculation results of multiple bands in the first remote sensing satellite image are obtained, wherein each band calculation result includes multiple parameter indices. An inversion dataset and a test dataset are randomly extracted from the soil moisture measurement data and the multiple band calculation results, respectively. The inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model. The inversion dataset and the multiple band calculation results are input into a full subset filtering algorithm to select the optimal variable combination. The optimal variable combination and the inversion dataset are input into multiple preset algorithm models to obtain the inversion model established by each algorithm. The test dataset is input into the inversion model established by each algorithm, and the inversion models established by the multiple algorithms are tested by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

[0189] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0190] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0191] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting regional soil moisture, characterized in that, include: Acquire current remote sensing satellite imagery and soil moisture sensor data within the target area; The current remote sensing satellite image is input into the latest target area soil moisture inversion model so that the latest target area soil moisture inversion model can invert the soil moisture data of each pixel in the target area; Based on the soil moisture sensor data, the soil moisture data of each pixel is corrected to obtain the soil moisture detection result of the target area; The soil moisture inversion model for the target area is trained based on the following steps: Acquire soil moisture measurement data in the sample soil and the first remote sensing satellite imagery for the corresponding measurement time period; Band calculations are performed on the first remote sensing satellite image to obtain multiple band calculation results reflected in the first remote sensing satellite image, wherein each band calculation result includes multiple parameter indices; Randomly extract an inversion dataset and a test dataset from the soil moisture measurement data and the multiple band calculation results, respectively. The inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model. The inversion dataset and the results of the multiple band operations are input into the full subset filtering algorithm to filter and obtain the optimal combination of variables. The optimal combination of variables and the inversion dataset are input into a variety of preset algorithm models to obtain the inversion model established by each algorithm. The test dataset is input into the inversion model established by each algorithm, and the inversion models established by the various algorithms are tested by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

2. The method according to claim 1, characterized in that, The step of correcting the soil moisture data based on the soil moisture sensor data to obtain the soil moisture detection result of the target area includes: The ratio of the soil moisture data of each pixel to the corresponding moisture sensor data is used as a correction coefficient; The ratio of the soil moisture data of each pixel to the correction coefficient is used as the soil moisture detection result of the target area; or, If no current remote sensing satellite imagery of the target area is acquired within a preset time period, the previously obtained soil moisture data is corrected based on the soil moisture sensor data.

3. The method according to claim 1, characterized in that, The acquisition of soil moisture measurement data in the sample soil includes: Based on preset rules, the number and location of soil sampling plots are determined within the target area, resulting in multiple soil sampling plots. Soil moisture measurement data were acquired using soil moisture sensors in the multiple soil sampling plots.

4. The method according to claim 1, characterized in that, The step of performing band calculations on the first remote sensing satellite image to obtain multiple band calculation results reflected in the first remote sensing satellite image includes: The reflectance in the blue band, green band, red band, and near-infrared band displayed in the first remote sensing satellite image is obtained. The calculation results of multiple bands reflected in the first remote sensing satellite image are calculated based on the blue light band reflectance, green light band reflectance, red light band reflectance and near-infrared band reflectance and the soil line coefficient of the sample soil.

5. The method according to claim 3 or 4, characterized in that, The random extraction of inversion datasets and test datasets from the soil moisture measurement data and the multiple band calculation results respectively includes: Based on a preset ratio and a preset number of extractions, inversion datasets and test datasets are randomly extracted from the soil moisture measurement data and the calculation results of the multiple bands, respectively, to obtain the inversion dataset and test dataset for each extraction.

6. The method according to claim 5, characterized in that, The step of inputting the inversion dataset and the results of the multiple band operations into a full subset filtering algorithm to filter for the optimal variable combination includes: The model is trained based on the inversion dataset and test dataset extracted each time; The goodness-of-fit of the model trained based on each extracted inversion dataset and test dataset is evaluated, and the optimal combination of variables with the best goodness-of-fit is selected.

7. The method according to claim 6, characterized in that, The process involves inputting the optimal variable combination and the inversion dataset into multiple preset algorithm models to obtain inversion models built using various algorithms, including: The optimal combination of variables and the inversion dataset extracted each time are input into a variety of preset algorithm models to obtain multiple inversion models established by each algorithm.

8. The method according to claim 7, characterized in that, The process involves inputting the test dataset into the inversion model established by each algorithm, and testing the inversion models established by the various algorithms by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area. The test dataset is input into multiple inversion models established by each algorithm, and the inversion models are verified based on goodness of fit and root mean square error. The inversion model with the highest goodness of fit and the smallest root mean square error is selected as the optimal soil moisture inversion model for the target area. Alternatively, the relative error coefficient of each inversion model can be calculated based on the goodness of fit and root mean square error, and the inversion model with the highest relative error coefficient can be selected as the optimal soil moisture inversion model for the target area.

9. A regional soil moisture detection device, characterized in that, include: The acquisition module is used to acquire current remote sensing satellite imagery and soil moisture sensor data within the target area; The detection module is used to input the current remote sensing satellite image into the latest target area soil moisture inversion model, so that the latest target area soil moisture inversion model can invert the soil moisture data of each pixel in the target area; The correction module is used to correct the soil moisture data of each pixel based on the soil moisture sensor data to obtain the soil moisture detection result of the target area; The model training module is used to acquire soil moisture measurement data and first remote sensing satellite images within the corresponding measurement time period; perform band calculations on the first remote sensing satellite images to obtain multiple band calculation results reflected in the first remote sensing satellite images, wherein each band calculation result includes multiple parameter indices; randomly extract an inversion dataset and a test dataset from the soil moisture measurement data and the multiple band calculation results, wherein the inversion dataset is used to train the soil moisture inversion model, and the test dataset is used to test the trained soil moisture inversion model; input the inversion dataset and the multiple band calculation results into a full subset filtering algorithm to filter out the optimal variable combination; input the optimal variable combination and the inversion dataset into a preset set of multiple algorithm models to obtain an inversion model established by each algorithm; input the test dataset into the inversion models established by each algorithm, and test the inversion models established by the multiple algorithms by calculating the coefficient of determination and root mean square error to determine the optimal soil moisture inversion model for the target area.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the regional soil moisture detection method as described in any one of claims 1 to 8.

Images