A soil moisture remote sensing estimation method, device, medium and terminal

Abstract

The application discloses a kind of soil moisture remote sensing estimation method, device, medium and terminal, the method includes: data acquisition is carried out to preset area, obtains the soil parameter and remote sensing detection parameter of the preset area;Model based on learning algorithm is constructed, the soil parameter and remote sensing detection parameter are input into the model based on learning algorithm and are trained, and quantitative estimation model of training is obtained;Remote sensing monitoring is carried out to preset area, and real-time monitoring data is obtained, and the real-time monitoring data is input into quantitative estimation model of training and is calculated, and soil moisture data is output to obtain;The application adopts the above method, and soil moisture in large-area land can be accurately estimated without collecting a large amount of soil data.

Description

Technical Field

[0001] This invention relates to the field of remote sensing estimation, and more particularly to a method, apparatus, medium, and terminal for remote sensing estimation of soil moisture. Background Technology

[0002] Soil moisture is a key hydrological indicator, generally referring to the absolute water content of the soil. Soil moisture control and its relationship with various water cycle processes include evaporation, transpiration, infiltration, and surface / groundwater runoff. A thorough understanding of soil moisture content and its spatiotemporal dynamics is crucial for agricultural water resource management. It not only allows us to understand crop water requirements but also maximizes support for scientific irrigation, minimizes ineffective water waste, and provides significant guidance for agricultural production.

[0003] Existing methods for estimating soil moisture require extensive soil sampling in the field, which is labor-intensive and cannot provide accurate estimates of soil moisture content. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this application is to provide a method, device, medium and terminal for remote sensing estimation of soil moisture, which aims to solve the problem of accurately estimating soil moisture content while reducing workload.

[0005] To address the aforementioned technical problems, the first aspect of this application provides a method for remote sensing estimation of soil moisture, the method comprising:

[0006] Data is collected from a preset area to obtain soil parameters and remote sensing detection parameters of the preset area;

[0007] A model based on a learning algorithm is constructed, and the soil parameters and remote sensing detection parameters are input into the model based on the learning algorithm for training to obtain a trained quantitative estimation model.

[0008] Remote sensing monitoring is performed on a preset area to obtain real-time monitoring data. The real-time monitoring data is then input into a trained quantitative estimation model for calculation, and soil moisture data is output.

[0009] As a further improved technical solution, the step of collecting data from a preset area to obtain soil parameters and remote sensing detection parameters of the preset area includes:

[0010] Soil samples were collected from the preset area, and the collected soil samples were measured and analyzed to obtain soil information.

[0011] The soil information is processed to obtain soil parameters in a spatial format;

[0012] Optical remote sensing is performed on the surface of the preset area to obtain optical remote sensing data. The optical remote sensing data is then processed to obtain remote sensing parameters.

[0013] As a further improved technical solution, the step of collecting soil samples from the preset area and analyzing the collected soil to obtain soil information includes:

[0014] Select several sampling points within the preset area and extract soil samples at the sampling points;

[0015] The soil samples were analyzed to obtain soil texture, dry bulk density, porosity and organic matter content;

[0016] Soil moisture characteristic curves were obtained by centrifugation of the soil samples, and soil saturated hydraulic conductivity, field capacity and permanent wilting coefficient were obtained from the soil moisture characteristic curves.

[0017] As a further improved technical solution, the soil information is processed to obtain spatially formatted soil parameters, including:

[0018] The soil texture, dry bulk density, porosity, organic matter content, soil saturated hydraulic conductivity, field water holding capacity, and permanent wilting coefficient were spatially characterized using thin plate spline interpolation to obtain spatially formatted soil parameters.

[0019] As a further improved technical solution, the step of performing optical remote sensing detection on the surface of the preset area to obtain optical remote sensing data, and processing the optical remote sensing data to obtain remote sensing detection parameters includes:

[0020] UAVs are used to collect optical remote sensing data, and the optical remote sensing data is preprocessed to obtain the reflectance of the R, G, B and NIR bands.

[0021] The EVI index is obtained by calculating the band reflectance of the R, G, B and NIR bands;

[0022] The remote sensing parameters include NIR band reflectance and EVI index.

[0023] As a further improved technical solution, the preprocessing of the optical remote sensing data includes:

[0024] Image stitching and georegistration are performed on optical remote sensing data to obtain a complete image;

[0025] The complete image is classified using the maximum natural classification method to remove shadows and dark pixels from the image.

[0026] As a further improved technical solution, the construction of a learning algorithm-based model involves inputting the soil parameters and remote sensing detection parameters into the learning algorithm-based model for training, resulting in a trained quantitative estimation model, including:

[0027] A hierarchical distribution modeling strategy was used to select modeling variables for the soil parameters and remote sensing detection parameters in the aforementioned spatial format;

[0028] A LightGBM model is constructed by selecting the LightGBM algorithm. The modeling variables are then input into the LightGBM model for calculation to obtain estimated values.

[0029] The observed true values ​​are input into the LightGBM model. The observed true values, combined with the estimated values, can be used to quantitatively analyze the model's performance. When the analysis indicators reach the predetermined values, a trained quantitative estimation model is obtained.

[0030] As a further improved technical solution, the acquisition of the observation truth value includes:

[0031] Sensors are installed in the soil within a preset area. The sensors detect soil moisture over time to obtain time-series soil moisture profile data, which is then used as the true observation value.

[0032] As a further improved technical solution, the step of remotely monitoring the preset area to obtain real-time monitoring data, inputting the real-time monitoring data into a trained quantitative estimation model for calculation, and outputting soil moisture data includes:

[0033] A drone is used to conduct real-time remote sensing monitoring of a preset area to obtain real-time monitoring data. The monitoring data is then input into a trained quantitative estimation model for calculation, and soil moisture data is output.

[0034] A second aspect of this application provides a soil moisture remote sensing estimation device, comprising:

[0035] The data acquisition module is used to acquire data from a preset area and obtain soil parameters and remote sensing detection parameters of the preset area.

[0036] The training module is used to build a model based on a learning algorithm. The soil parameters and remote sensing detection parameters are input into the model based on the learning algorithm for training, and a trained quantitative estimation model is obtained.

[0037] The calculation module is used to remotely monitor a preset area, obtain real-time monitoring data, input the real-time monitoring data into a trained quantitative estimation model for calculation, and output soil moisture data.

[0038] A third aspect of this application provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in any of the above-described soil moisture remote sensing estimation methods.

[0039] A fourth aspect of this application provides a terminal device, which includes: a processor, a memory, and a communication bus; the memory stores a computer-readable program that can be executed by the processor;

[0040] The communication bus enables communication between the processor and the memory;

[0041] When the processor executes the computer-readable program, it implements the steps in any of the above-described soil moisture remote sensing estimation methods.

[0042] Beneficial effects: Compared with existing technologies, the soil moisture remote sensing estimation method of the present invention includes: collecting data from a preset area to obtain soil parameters and remote sensing detection parameters of the preset area; constructing a model based on a learning algorithm, inputting the soil parameters and remote sensing detection parameters into the model based on the learning algorithm for training, and obtaining a trained quantitative estimation model; performing remote sensing monitoring on the preset area to obtain real-time monitoring data, inputting the real-time monitoring data into the trained quantitative estimation model for calculation, and outputting soil moisture data; the present invention, by adopting the above method, can accurately estimate soil moisture in a large area of ​​land without collecting a large amount of soil data. Attached Figure Description

[0043] Figure 1 This is a flowchart of the soil moisture remote sensing estimation method of the present invention.

[0044] Figure 2 This is a framework diagram of model training in the soil moisture remote sensing estimation method of the present invention.

[0045] Figure 3 This is a structural schematic diagram of the terminal device provided by the present invention.

[0046] Figure 4 This is a structural diagram of the soil moisture remote sensing estimation device provided by the present invention.

[0047] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0048] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings. Preferred embodiments of this application are shown in the drawings. However, this application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0050] The inventors discovered through research that the existing technology has the following problems:

[0051] (1) Currently, when using remote sensing to monitor soil moisture, satellites are typically used. However, the spatial resolution of soil moisture data obtained by microwave satellite remote sensing is relatively coarse, mostly tens of kilometers, which greatly limits its applicability in fields and irrigation areas. Compared with microwave remote sensing, optical and thermal infrared satellite remote sensing have higher resolution and can be used to measure soil moisture at medium resolution scales. However, medium resolution scales are still relatively coarse for precise irrigation management in fields and cannot meet the actual production needs. In addition, due to the limited penetration depth of optical and thermal infrared radiation, the corresponding sensors can usually only capture changes in soil moisture close to the surface, and light and heat signals cannot penetrate clouds, which results in limited observation time and makes it difficult to generate continuous time series satellite observation data.

[0052] (2) In addition, existing soil moisture estimates generally only focus on surface (2cm) moisture and do not estimate near-surface (10cm) and root zone (50cm) moisture, which affects precise irrigation management in the field.

[0053] To address the aforementioned problems, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0054] like Figure 1 As shown in the embodiment of this application, a method for remote sensing estimation of soil moisture includes the following steps:

[0055] S1, Data is collected from the preset area to obtain the soil parameters and remote sensing detection parameters of the preset area;

[0056] Specifically, the preset area is a region where soil moisture will be estimated. Soil parameters can be obtained through soil sampling, measurement and analysis, and remote sensing parameters can be obtained through drone remote sensing monitoring.

[0057] The step of collecting data from a preset area to obtain soil parameters and remote sensing detection parameters for that preset area includes:

[0058] S101, collect soil samples from the preset area, measure and analyze the collected soil samples, and obtain soil information;

[0059] The step of collecting soil samples from the preset area and analyzing the collected soil to obtain soil information includes:

[0060] Select several sampling points within the preset area and extract soil samples at the sampling points;

[0061] The soil samples were analyzed to obtain soil texture, dry bulk density, porosity and organic matter content;

[0062] Soil moisture characteristic curves were obtained by centrifugation of the soil samples, and soil saturated hydraulic conductivity, field capacity and permanent wilting coefficient were obtained from the soil moisture characteristic curves.

[0063] Specifically, soil saturated hydraulic conductivity, field capacity, and permanent wilting coefficient are important soil hydraulic parameters. Soil saturated hydraulic conductivity affects precipitation infiltration and runoff patterns, thus influencing regional water redistribution. Field capacity is the maximum water content that saturated soil can retain after gravity drainage, representing the upper limit of soil water available to most plants. The permanent wilting coefficient is the soil water content at which plants exhibit permanent wilting, determining the lower limit of soil water available to plants. All three are also crucial sensitive parameters in eco-hydrological models, determining farmland water requirements and irrigation regimes in agricultural practice. Soil physical properties control the spatial variability and temporal dynamics of soil moisture; assimilating these properties into a soil water transport function allows for statistical estimation of soil water capacity and saturated hydraulic conductivity parameters.

[0064] During implementation, a preset area is first set up, and 50 sampling points are selected within the preset area according to the corresponding sampling method. The corresponding sampling method is to select sampling points from the total set by comprehensively considering the heterogeneity of land use and land cover type, topography and soil type in the area and using the CLHS (Latin Hypercube Sampling) method to select candidate points.

[0065] After determining the sampling points, use a soil auger to extract the corresponding number of complete soil column samples at the sampling points. The soil column is 50 cm high and 5 cm in diameter. Seal the soil column and prepare for relevant experimental analysis.

[0066] The soil texture, dry bulk density, porosity, and organic matter content of the collected soil samples were determined. Soil texture includes the content of clay, silt, and sand.

[0067] The determination process was as follows: Undisturbed soil samples were dried in an oven at 105℃ to constant weight, then weighed to calculate soil bulk density. Disturbed soil samples were air-dried and pretreated by passing them through 0.25mm and 1mm mesh sieves for determining soil organic carbon content and particle size distribution. Soil organic matter content was determined using the potassium dichromate oxidation method, and soil particle size distribution was determined using a laser particle size analyzer. Total soil porosity was calculated based on the relationship between soil bulk density and soil density (a constant value of 2.65 g / cm³). From these steps, soil texture, dry bulk density, porosity, and organic matter content were obtained.

[0068] Three indicators were selected to assess soil hydraulic properties: field water holding capacity, permanent wilting coefficient, and saturated hydraulic conductivity.

[0069] The soil samples were measured using a centrifuge method and fitted with the van Genuchten model to obtain soil moisture characteristic curves. Soil saturated hydraulic conductivity, field capacity, and permanent wilting coefficient were obtained from the soil moisture characteristic curves.

[0070] S102, The soil information is processed to obtain soil parameters in spatial format;

[0071] The spatial format soil parameters obtained by processing the soil information include:

[0072] The soil texture, dry bulk density, porosity, organic matter content, soil saturated hydraulic conductivity, field water holding capacity, and permanent wilting coefficient were spatially characterized using thin plate spline interpolation to obtain spatially formatted soil parameters.

[0073] Specifically, soil texture, dry bulk density, porosity, organic matter content, saturated hydraulic conductivity, field capacity, and permanent wilting coefficient are all single-point data obtained through ground sampling. Thin-plate spline interpolation is used to spatially represent these data. The spline function method employs an interpolation technique that estimates values ​​using a mathematical function that minimizes the total surface curvature, thereby generating a smooth surface that passes precisely through the input points. Specifically, regular spline function and tension spline function methods are weighted and combined to maximize the spatial fit of the soil property data. The regular spline function method uses values ​​that may lie outside the range of the sample data to create a gradually smooth surface. The tension spline function method controls the surface hardness based on the characteristics of the modeled phenomenon.

[0074] S103, perform optical remote sensing detection on the surface of the preset area to obtain optical remote sensing data, and process the optical remote sensing data to obtain remote sensing detection parameters.

[0075] The step of performing optical remote sensing detection on the surface of the preset area to obtain optical remote sensing data, and processing the optical remote sensing data to obtain remote sensing detection parameters includes:

[0076] UAVs are used to collect optical remote sensing data, and the optical remote sensing data is preprocessed to obtain the reflectance of the R, G, B and NIR bands.

[0077] The EVI index is obtained by calculating the band reflectance of the R, G, B and NIR bands;

[0078] The remote sensing parameters include NIR band reflectance and EVI index.

[0079] The preprocessing of the optical remote sensing data includes:

[0080] Image stitching and georegistration are performed on optical remote sensing data to obtain a complete image;

[0081] The complete image is classified using the maximum natural classification method to remove shadows and dark pixels from the image.

[0082] Specifically, a DJI drone platform equipped with a multispectral camera was used to collect optical remote sensing data, ensuring that the horizontal and vertical overlap of the collected images reached 80%. After drone data acquisition, Pix4d Mapper Pro software was used to preprocess the data, performing image stitching and georegistration on the optical remote sensing data to obtain a complete image. To reduce the uncertainty of soil moisture retrieval, the complete image needed to be classified in advance using the maximum natural classification method to remove shadows and dark pixels in the image. Finally, the reflectance of the R, G, B, and NIR bands was obtained. The EVI index was obtained through the Enhanced Vegetation Index (EVI) algorithm, as follows:

[0083]

[0084] In the formula, ρ NIR ρ is the band reflectance of NIR. RED ρ is the reflectivity of the R-band. BLUE For B-band reflectivity;

[0085] The EVI algorithm can simultaneously reduce the impact of atmospheric and soil noise, stably reflecting the vegetation conditions of the measured area. The narrower range settings of the red and near-infrared detection bands not only improve the ability to detect sparse vegetation but also reduce the influence of water vapor. At the same time, the blue light band is introduced to correct for atmospheric aerosol scattering and soil background, ultimately yielding NIR band reflectance and EVI index.

[0086] S2, Construct a model based on a learning algorithm, input the soil parameters and remote sensing detection parameters into the model based on the learning algorithm for training, and obtain a trained quantitative estimation model;

[0087] Specifically, in this embodiment, the learning algorithm-based model is the LightGBM model. However, the model can also be other learning algorithm-based models. The soil parameters and remote sensing detection parameters mentioned above are input into the LightGBM model for training, and finally a trained quantitative estimation model is obtained.

[0088] The step of constructing a model based on a learning algorithm involves inputting the soil parameters and remote sensing detection parameters into the model for training to obtain a trained quantitative estimation model, including:

[0089] S201, a hierarchical distribution modeling strategy is used to select modeling variables for the soil parameters and remote sensing detection parameters in the spatial format;

[0090] S202, a LightGBM model is constructed by selecting the LightGBM algorithm, and the modeling variables are input into the LightGBM model for calculation to obtain the estimated value;

[0091] S203, the observed true values ​​are input into the LightGBM model. The observed true values ​​combined with the estimated values ​​can be used to quantitatively analyze the model performance. When the analysis index reaches the predetermined value, the trained quantitative estimation model is obtained.

[0092] The acquisition of the observed truth value includes:

[0093] Sensors are installed in the soil within a preset area. The sensors detect soil moisture over time to obtain time-series soil moisture profile data, which is then used as the true observation value.

[0094] To train the machine learning model and evaluate the inversion performance of surface, near-surface, and root zone soil moisture, five TDR sensors were installed in the soil of a preset area at burial depths of 2 cm, 10 cm, and 50 cm, corresponding to the surface, near-surface, and root zone, respectively, to better reflect the spatial distribution of soil moisture. The data measurement step was 30 minutes, and the measured data were used as the true observation values.

[0095] A hierarchical distribution modeling strategy was adopted to select modeling variables for spatial format soil parameters and remote sensing detection parameters. The hierarchical distribution grouped the spatial format soil parameters and remote sensing detection parameters according to the surface, near-surface, and root zone, and then set multiple covariate groups C91(9), C92(36), C93(84), C94(126), C95(126), C97(36), C98(9), and C99(1) in a stepwise arrangement and combination. The hierarchical strategy was used to construct multiple different models for machine learning algorithms to estimate soil moisture in the surface, near-surface, and root zone. Li was selected. The ghtGBM algorithm for building the lightGBM model comprises two key elements: "light" signifies lightweight, and "GBM gradient boosting machine." LightGBM is a gradient boosting framework that uses a decision tree based on a learning algorithm. Multiple covariate sets are input into the LightGBM model for computation to obtain estimated values. To evaluate the performance of the LightGBM model, soil moisture measured by TDR is used as the observed true value. Combined with the estimated values, the root mean square error (RMSE), Pearson correlation coefficient (R), and quartile relative prediction error (RPIQ) are used to quantitatively analyze the model's performance. Once the analysis indicators reach predetermined values, the trained quantitative estimation model is obtained.

[0096] The LightGBM model needs to calculate the relationship between soil saturated hydraulic conductivity, field capacity, and permanent wilting coefficient across different soil layers. For example, the formula: y = a·x + b

[0097] The calculation ultimately achieves the goal of deducing the corresponding parameters in the deeper layers using only surface data. In the formula, y is the saturated hydraulic conductivity of the third soil layer, X is the saturated hydraulic conductivity that can be measured in the surface layer, b is a constant term, and a is a coefficient.

[0098] S3. Perform remote sensing monitoring on the preset area to obtain real-time monitoring data. Input the real-time monitoring data into the trained quantitative estimation model for calculation and output soil moisture data.

[0099] Specifically, after training the above model, real-time remote sensing monitoring can be carried out on the preset area. The obtained real-time detection data is input into the trained quantitative estimation model for calculation, and finally the soil moisture data is output.

[0100] The process of remotely monitoring a preset area to obtain real-time monitoring data, inputting the real-time monitoring data into a trained quantitative estimation model for calculation, and outputting soil moisture data includes:

[0101] S301, using a drone to perform real-time remote sensing monitoring of a preset area, obtaining real-time monitoring data, inputting the monitoring data into a trained quantitative estimation model for calculation, and outputting soil moisture data.

[0102] Specifically, unmanned aerial vehicles (UAVs) are used to conduct real-time remote sensing monitoring of a pre-defined area to obtain real-time monitoring data. This real-time monitoring data is real-time land optical remote sensing data. The real-time monitoring data is input into a trained quantitative estimation model for calculation, and the calculation results are processed using median filtering and used as the true value for inversion. Specifically, since the acquired remote sensing data is in pixel form, considering the proximity effect of pixels, median filtering is used to set the gray value of each pixel to the median of the gray values ​​of all pixels within a certain neighborhood window of that pixel, and this median is used as the true value for final inversion. This yields soil moisture data at different depth levels, such as surface, near-surface, and root zone soil moisture data. By obtaining surface soil moisture, near-surface and root zone soil moisture data can be deduced.

[0103] Practical application examples:

[0104] like Figure 2 As shown, model training, in practice, includes the following steps:

[0105] A1, Obtain soil data for the preset area;

[0106] A2, acquire remote sensing data for the preset area;

[0107] A3. The obtained spatial format soil property data, NIR band reflectance and EVI spectral index are used to screen out modeling variables using a hierarchical distribution modeling strategy.

[0108] A4. Input the selected modeling variables into the LightGBM model;

[0109] A5 is used for model training.

[0110] Soil data acquisition specifically includes the following steps:

[0111] A101, CHLS soil sampling;

[0112] A102, Physicochemical Experimental Analysis of Soil Parameters;

[0113] A103 yielded soil texture, dry bulk density, porosity, organic matter content, field water holding capacity, permanent wilting coefficient, and saturated hydraulic conductivity.

[0114] A104, optimize the parameters in step A103 using thin-disk spline interpolation;

[0115] A105 yields spatial format soil property data.

[0116] The acquisition of remote sensing data specifically includes the following steps:

[0117] A201 uses drones to collect remote sensing images of a pre-defined area;

[0118] A202, stitching together the acquired remote sensing images;

[0119] A203, Processing and removing shadows from the stitched remote sensing images;

[0120] A204 yields the NIR band reflectance and EVI spectral index.

[0121] Model training includes:

[0122] A501 inputs time-series soil moisture profile data obtained from TDR sensors into the training model;

[0123] A502, accuracy verification: compare the time-series soil moisture profile data obtained by the TDR sensor with the calculation results. If the soil moisture profile data is different from the calculation results of the training model, recalculate.

[0124] A503, Output Results: If the soil moisture profile data matches the calculation results of the training model, then output the results.

[0125] This invention, employing the aforementioned method, considers soil physical and hydraulic properties and combines UAV remote sensing data for soil moisture estimation. It overcomes the limitations of single sensors, which typically only capture near-surface soil moisture changes and have low temporal resolution, making it difficult to generate continuous time-series satellite observation data. The CLHS (Latin Hypercube Sampling) method is used to select candidate points from the overall dataset, improving sample point representativeness. Basic preprocessing is performed after obtaining the remote sensing image data, using the EVI index as a representative, to reduce input noise and overcome the problems of vegetation index saturation and lack of linear relationship with actual vegetation cover. The data were used as input to the model; thin-plate spline interpolation was used to spatially represent soil physical and hydraulic properties and apply them to soil moisture estimation; variable groups were selected hierarchically in a stepwise arrangement and combination manner to enrich the basic data for model input; combined with the selected variables, the lightGBM model was used to perform packaged estimation of soil moisture at the surface (2cm), near-surface (10cm), and root zone (50cm); after completing model training and estimation, the median filtering method was used to process the results and treat them as the true values ​​for inversion; relying on the above techniques, the accuracy of soil moisture inversion was improved. Using the above method, soil moisture in large areas of land can be accurately estimated without collecting a large amount of soil data. Regional soil moisture content can be obtained through inversion, allowing for timely understanding of soil moisture conditions and assessment of whether crop water supply is sufficient.

[0126] like Figure 4 As shown, based on the above-mentioned soil moisture remote sensing estimation method, this embodiment provides a soil moisture remote sensing estimation device, including:

[0127] The data acquisition module is used to acquire data from a preset area and obtain soil parameters and remote sensing detection parameters of the preset area.

[0128] The training module is used to build a model based on a learning algorithm. The soil parameters and remote sensing detection parameters are input into the model based on the learning algorithm for training, and a trained quantitative estimation model is obtained.

[0129] The calculation module is used to remotely monitor a preset area, obtain real-time monitoring data, input the real-time monitoring data into a trained quantitative estimation model for calculation, and output soil moisture data.

[0130] Furthermore, it is worth noting that the working process of the soil moisture remote sensing estimation device provided in this embodiment is the same as that of the soil moisture remote sensing estimation method described above. For details, please refer to the working process of the soil moisture remote sensing estimation method, which will not be repeated here.

[0131] Based on the above-described soil moisture remote sensing estimation method, this embodiment provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the soil moisture remote sensing estimation method as described in the above embodiment.

[0132] Based on the above-mentioned remote sensing estimation method for soil moisture, this application also provides a terminal device, such as... Figure 3 As shown, it includes at least one processor 20; a display screen 21; and a memory 22, and may also include a communications interface 23 and a bus 24. The processor 20, display screen 21, memory 22, and communications interface 23 can communicate with each other via the bus 24. The display screen 21 is configured to display a preset user guide interface in the initial setup mode. The communications interface 23 can transmit information. The processor 20 can invoke logical instructions in the memory 22 to execute the methods described in the above embodiments.

[0133] Furthermore, the logical instructions in the aforementioned memory 22 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0134] The memory 22, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, such as program instructions or modules corresponding to the methods in the embodiments of this disclosure. The processor 20 executes functional applications and data processing by running the software programs, instructions, or modules stored in the memory 22, thereby implementing the methods in the above embodiments.

[0135] The memory 22 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 22 may include high-speed random access memory (RAM) and non-volatile memory. Examples include 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, as well as transient storage media.

[0136] By adopting the above method, this invention can accurately estimate soil moisture in a large area of ​​land without collecting a large amount of soil data. By inversion, the regional soil moisture content can be obtained, the soil moisture situation can be grasped in a timely manner, and the water supply for crops can be judged as sufficient.

[0137] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method of soil moisture remote sensing estimation, characterized by, The method comprises: data collection on a preset area to obtain soil parameters and remote sensing detection parameters of the preset area, the soil parameters being soil parameters in a spatial format, and the remote sensing detection parameters including NIR band reflectivity and EVI index; constructing a model based on a learning algorithm, inputting the soil parameters and remote sensing detection parameters into the model based on the learning algorithm for training to obtain a trained quantitative estimation model; remote sensing monitoring on the preset area to obtain real-time monitoring data, inputting the real-time monitoring data into the trained quantitative estimation model for calculation to output soil moisture data; the constructing of the model based on the learning algorithm, the inputting of the soil parameters and remote sensing detection parameters into the model based on the learning algorithm for training to obtain the trained quantitative estimation model comprises: adopting a layered distribution modeling strategy to screen modeling variables from the soil parameters in the spatial format and the remote sensing detection parameters; constructing a LightGBM model by selecting a LightGBM algorithm, inputting the modeling variables into the LightGBM model for calculation to obtain estimation values; inputting observation true values into the LightGBM model, the observation true values in combination with the estimation values can quantitatively analyze the effect of the model, and the analysis index reaches a predetermined value to obtain the trained quantitative estimation model; the obtaining of the observation true values comprises: installing sensors in the soil of the preset area, the sensors detecting soil moisture in a time sequence to obtain time sequence soil moisture profile data, and taking the time sequence soil moisture profile data as the observation true values; the sensors are installed on the ground surface, near the ground surface and in the root zone.

2. The method of claim 1, wherein, the data collection on the preset area to obtain the soil parameters and remote sensing detection parameters of the preset area comprises: collecting soil in the preset area, and determining and analyzing the collected soil to obtain soil information; processing the soil information to obtain soil parameters in a spatial format; optically remotely sensing the ground surface in the preset area to obtain optical remote sensing data, and processing the optical remote sensing data to obtain remote sensing detection parameters.

3. The method of claim 2, wherein, the collecting of the soil in the preset area, the determining and analyzing of the collected soil to obtain soil information comprises: selecting a plurality of sample points in the preset area, and extracting soil samples at the sample points; determining and analyzing the soil samples to obtain soil texture, dry bulk density, porosity and organic matter content; determining soil moisture characteristic curves of the soil samples by a centrifuge method to obtain soil saturated hydraulic conductivity, field capacity and permanent wilting coefficient from the soil moisture characteristic curves.

4. The method of claim 3, wherein, the processing of the soil information to obtain soil parameters in a spatial format comprises: spatially characterizing the soil texture, dry bulk density, porosity, organic matter content, soil saturated hydraulic conductivity, field capacity and permanent wilting coefficient by a thin plate spline interpolation method to obtain soil parameters in a spatial format.

5. The method of claim 4, wherein, the optically remotely sensing of the ground surface in the preset area to obtain optical remote sensing data, and the processing of the optical remote sensing data to obtain remote sensing detection parameters comprises: UAVs are used to collect optical remote sensing data, and the optical remote sensing data is preprocessed to obtain the reflectance of the R, G, B and NIR bands. The EVI index is obtained by calculating the reflectivity of the R, G, B and NIR bands.

6. The method of claim 5, wherein, Preprocessing the optical remote sensing data includes: Image stitching and georegistration are performed on optical remote sensing data to obtain a complete image; The complete image is classified using the maximum natural classification method to remove shadows and dark pixels from the image.

7. The method of claim 1, wherein, The process of remotely monitoring a preset area to obtain real-time monitoring data, inputting the real-time monitoring data into a trained quantitative estimation model for calculation, and outputting soil moisture data includes: A drone is used to conduct real-time remote sensing monitoring of a preset area to obtain real-time monitoring data. The monitoring data is then input into a trained quantitative estimation model for calculation, and soil moisture data is output.

8. A device for soil moisture remote sensing estimation, for implementing the steps of the method for soil moisture remote sensing estimation according to any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire data from a preset area and obtain soil parameters and remote sensing detection parameters of the preset area. The training module is used to build a model based on a learning algorithm. The soil parameters and remote sensing detection parameters are input into the model based on the learning algorithm for training, and a trained quantitative estimation model is obtained. The calculation module is used to remotely monitor a preset area, obtain real-time monitoring data, input the real-time monitoring data into a trained quantitative estimation model for calculation, and output soil moisture data.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the soil moisture remote sensing estimation method as described in any one of claims 1-7.

10. A terminal device, comprising: include: Processor, memory, and communication bus; the memory stores a computer-readable program that can be executed by the processor; The communication bus enables communication between the processor and the memory; When the processor executes the computer-readable program, it implements the steps in the soil moisture remote sensing estimation method as described in any one of claims 1-7.

Images