Soil moisture data downscaling method, device, electronic device and storage medium

Abstract

The present application relates to a soil moisture downscaling method, device, electronic device and storage medium. The method trains a diffusion model by performing forward diffusion and reverse denoising processes on soil moisture residual data between interpolated soil moisture data with high spatial resolution and real soil moisture data with high spatial resolution, so that the diffusion model can predict soil moisture residual data from standard Gaussian noise. Based on the interpolated soil moisture data with high spatial resolution and the predicted soil moisture residual data, downscaled soil moisture data is obtained, and the downscaling of soil moisture data is achieved. The spatial resolution of the downscaled soil moisture data is high, which can meet the needs of refined research and application.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a soil moisture data downscaling method, device, electronic device and storage medium. Background Art

[0002] Soil moisture is a key parameter in the fields of hydrology, agriculture and climate. Soil moisture is essential for plant growth, ecosystem balance and climate regulation. It is directly related to crop yield, soil microbial activity, nutrient cycling and soil structure. Monitoring and managing soil moisture is of great significance for improving agricultural production efficiency, protecting the environment and responding to climate change.

[0003] However, the spatial resolution of existing soil moisture data is usually low, which makes it difficult to fully reflect the detailed spatial characteristics of regional soil moisture and cannot meet the needs of refined research and application. Summary of the invention

[0004] Based on this, the purpose of the present application is to provide a soil moisture data downscaling method, device, electronic device and storage medium, which can improve the spatial resolution of soil moisture data.

[0005] According to a first aspect of an embodiment of the present application, a soil moisture data downscaling method is provided, comprising the following steps:

[0006] Acquire soil moisture data of a first spatial resolution, latitude and longitude information and time information of the soil moisture data of the first spatial resolution, and standard Gaussian noise of a target time step;

[0007] Interpolating the soil moisture data of the first spatial resolution to obtain soil moisture data of the second spatial resolution; interpolating the longitude and latitude information of the soil moisture data of the first spatial resolution to obtain the longitude and latitude information of the soil moisture data of the second spatial resolution; the second spatial resolution is higher than the first spatial resolution;

[0008] Acquire vegetation coverage data, surface temperature data, and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data at the second spatial resolution;

[0009] Establish and train a diffusion model to obtain a trained diffusion model; wherein the diffusion model includes a forward diffusion module and a reverse diffusion module, and the reverse diffusion module includes a noise prediction module and a denoising module; Gaussian noise is gradually added to the sample soil moisture residual data through the forward diffusion module to obtain noisy soil moisture residual data at different time steps; the noise prediction module is used to perform noise prediction on the noisy soil moisture residual data at different time steps to obtain predicted noise; the noise prediction module is iteratively trained through the added Gaussian noise and the predicted noise to obtain a trained diffusion model;

[0010] Input the target time step, standard Gaussian noise of the target time step, soil moisture data of the second spatial resolution, time information and longitude and latitude information, vegetation coverage data, surface temperature data and soil texture data into the noise prediction module and denoising module of the trained diffusion model to obtain predicted soil moisture residual data;

[0011] The downscaled soil moisture data are obtained based on the predicted soil moisture residual data and the soil moisture data at the second spatial resolution.

[0012] According to a second aspect of an embodiment of the present application, a soil moisture data downscaling device is provided, comprising:

[0013] A soil moisture data acquisition module, used to acquire soil moisture data of a first spatial resolution, latitude and longitude information and time information of the soil moisture data of the first spatial resolution, and standard Gaussian noise of a target time step;

[0014] A soil moisture data interpolation module, configured to interpolate soil moisture data of a first spatial resolution to obtain soil moisture data of a second spatial resolution; interpolate longitude and latitude information of soil moisture data of the first spatial resolution to obtain longitude and latitude information of soil moisture data of the second spatial resolution; the second spatial resolution is higher than the first spatial resolution;

[0015] An auxiliary data acquisition module, used to acquire vegetation coverage data, surface temperature data and soil texture data having the same latitude and longitude information as the soil moisture data of the second spatial resolution;

[0016] A diffusion model training module is used to establish and train a diffusion model to obtain a trained diffusion model; wherein the diffusion model includes a forward diffusion module and a reverse diffusion module, and the reverse diffusion module includes a noise prediction module and a denoising module; Gaussian noise is gradually added to the sample soil moisture residual data through the forward diffusion module to obtain the noisy soil moisture residual data at different time steps; the noise prediction module is used to perform noise prediction on the noisy soil moisture residual data at different time steps to obtain the predicted noise; the noise prediction module is iteratively trained through the added Gaussian noise and the predicted noise to obtain the trained diffusion model;

[0017] A soil moisture residual data prediction module is used to input the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module and the denoising module of the trained diffusion model to obtain the predicted soil moisture residual data;

[0018] The soil moisture data downscaling module is used to obtain downscaled soil moisture data according to the predicted soil moisture residual data and the soil moisture data of the second spatial resolution.

[0019] According to a third aspect of an embodiment of the present application, there is provided an electronic device, comprising: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the steps of the method of the first aspect.

[0020] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the steps of the method of the first aspect are implemented.

[0021] The embodiment of the present application obtains soil moisture data of a first spatial resolution, the latitude and longitude information and time information of the soil moisture data of the first spatial resolution, and standard Gaussian noise of a target time step; interpolates the soil moisture data of the first spatial resolution to obtain soil moisture data of a second spatial resolution; interpolates the latitude and longitude information of the soil moisture data of the first spatial resolution to obtain the latitude and longitude information of the soil moisture data of the second spatial resolution; the second spatial resolution is higher than the first spatial resolution; obtains vegetation coverage data, surface temperature data, and soil texture data whose latitude and longitude information are the same as the latitude and longitude information of the soil moisture data of the second spatial resolution; establishes and trains a diffusion model to obtain a trained diffusion model; wherein the diffusion model includes a forward diffusion module and a reverse diffusion module, and the reverse diffusion module includes a noise prediction module. The invention relates to a method for determining the soil moisture residual data of a soil sample by using a forward diffusion module and a denoising module; the method further comprises the following steps: gradually adding Gaussian noise to the sample soil moisture residual data through a forward diffusion module to obtain the noisy soil moisture residual data at different time steps; performing noise prediction on the noisy soil moisture residual data at different time steps through a noise prediction module to obtain the predicted noise; iteratively training the noise prediction module through the added Gaussian noise and the predicted noise to obtain a trained diffusion model; inputting the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module and the denoising module of the trained diffusion model to obtain the predicted soil moisture residual data; obtaining the downscaled soil moisture data according to the predicted soil moisture residual data and the soil moisture data of the second spatial resolution. The embodiment of the present application first interpolates the soil moisture data of low spatial resolution to obtain the interpolated soil moisture data of high spatial resolution. Through the forward diffusion and reverse denoising process of the soil moisture residual data between the interpolated soil moisture data with high spatial resolution and the real soil moisture data with high spatial resolution, the diffusion model is trained so that the diffusion model can predict the soil moisture residual data from the standard Gaussian noise. Based on the interpolated soil moisture data with high spatial resolution and the predicted soil moisture residual data, the downscaled soil moisture data is obtained, and the downscaling of the soil moisture data is realized. The spatial resolution of the downscaled soil moisture data is high, which can meet the needs of refined research and application.

[0022] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application.

[0023] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1A schematic diagram of a process flow of a soil moisture data downscaling method provided in one embodiment of the present application;

[0025] Figure 2 A structural block diagram of a soil moisture data downscaling device provided in one embodiment of the present application;

[0026] Figure 3 A schematic block diagram of the structure of an electronic device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the objectives, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0028] It should be clear that the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0029] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the embodiments of the present application. The singular forms of "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0030] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims. In the description of the present application, it should be understood that the terms "first", "second", "third", etc. are only used to distinguish similar objects, and do not have to be used to describe a specific order or sequence, nor can they be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0031] In addition, in the description of this application, unless otherwise specified, "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0032] See also Figure 1 , which is a flow chart of a soil moisture data downscaling method provided in an embodiment of the present application. The soil moisture data downscaling method provided in an embodiment of the present application comprises the following steps:

[0033] S10: Acquire soil moisture data of a first spatial resolution, longitude and latitude information and time information of the soil moisture data of the first spatial resolution, and standard Gaussian noise of a target time step.

[0034] The time information refers to the time when the soil moisture data is captured or recorded. For example, the time information is January 1, 2023. The longitude and latitude information can reflect the spatial location of the soil moisture data.

[0035] Among them, standard Gaussian noise refers to random noise whose probability density function follows a standard normal distribution. The time step is represented by time t, and the target time step is t=T. The value of T can be set manually according to actual needs. For example, T=1000.

[0036] In an embodiment of the present application, soil moisture data of a first spatial resolution is obtained from a soil moisture product SMAP (soil moisture active passive). Specifically, the first spatial resolution is 9 km. The soil moisture product SMAP records the time information and longitude and latitude information of the soil moisture data, and the time information and longitude and latitude information of the corresponding soil moisture data can be extracted from the soil moisture product SMAP. The standard Gaussian noise of the target time step refers to the standard Gaussian noise sampled at the target time step, which can be expressed as , .

[0037] S20: interpolate the soil moisture data of the first spatial resolution to obtain soil moisture data of the second spatial resolution; interpolate the longitude and latitude information of the soil moisture data of the first spatial resolution to obtain the longitude and latitude information of the soil moisture data of the second spatial resolution; the second spatial resolution is higher than the first spatial resolution.

[0038] In an embodiment of the present application, a preset interpolation algorithm is used to interpolate soil moisture data of a first spatial resolution to obtain soil moisture data of a second spatial resolution. A preset interpolation algorithm is used to interpolate longitude and latitude information of soil moisture data of a first spatial resolution to obtain longitude and latitude information of soil moisture data of a second spatial resolution. Specifically, the second spatial resolution is 3 km. The preset interpolation algorithm includes but is not limited to nearest neighbor interpolation, bilinear interpolation, and bicubic interpolation algorithms.

[0039] S30: Acquire vegetation coverage data, surface temperature data, and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data at the second spatial resolution.

[0040] Among them, the vegetation coverage (Fractional Vegetation Cover, FVC) data refers to the percentage of the vertical projection area of ​​vegetation (including leaves, stems, and branches) on the ground to the total area of ​​the statistical area. The land surface temperature (LST) data refers to the temperature of the ground. After the sun's heat energy is radiated to the ground, part of it is reflected and part of it is absorbed by the ground, causing the ground to heat up. The temperature obtained after measuring the ground temperature is the surface temperature. Soil texture data refers to the combination of mineral particles (sand, silt, clay) of different sizes and diameters in the soil.

[0041] In an embodiment of the present application, after obtaining the longitude and latitude information of the soil moisture data of the second spatial resolution, vegetation coverage data, surface temperature data and soil texture data having the same longitude and latitude information as the soil moisture data of the second spatial resolution can be matched from the vegetation coverage data, surface temperature data and soil texture data.

[0042] S40: Establish and train a diffusion model to obtain a trained diffusion model; wherein the diffusion model includes a forward diffusion module and a reverse diffusion module, and the reverse diffusion module includes a noise prediction module and a denoising module; Gaussian noise is gradually added to the sample soil moisture residual data through the forward diffusion module to obtain noisy soil moisture residual data at different time steps; noise prediction is performed on the noisy soil moisture residual data at different time steps through the noise prediction module to obtain predicted noise; the noise prediction module is iteratively trained through the added Gaussian noise and the predicted noise to obtain a trained diffusion model.

[0043] Among them, the forward diffusion module is the first step of the diffusion model, and its main task is to gradually transform the original data into noise. Specifically, the forward diffusion process refers to the process of gradually adding Gaussian noise from the original data until the data is completely transformed into standard Gaussian noise. The reverse diffusion module is the second step of the diffusion model, including the noise prediction module and the denoising module. Its main task is to restore the original data by gradually removing the noise starting from the standard Gaussian noise.

[0044] In an embodiment of the present application, the forward diffusion module uses the sample soil moisture residual data as initial data, gradually adds Gaussian noise according to the set noise addition schedule, and obtains noisy soil moisture residual data at different time steps.

[0045] The noise prediction module predicts the Gaussian noise added between the initial data and the noisy soil moisture residual data at different time steps. The Gaussian noise added between the initial data and the noisy soil moisture residual data at different time steps is used as a label, and the noise prediction module can be trained based on the difference between the predicted Gaussian noise and the label.

[0046] The denoising module gradually denoises the noisy soil moisture residual data at different time steps to obtain the initial data.

[0047] S50: Input the target time step, standard Gaussian noise of the target time step, soil moisture data of the second spatial resolution, time information and longitude and latitude information, vegetation coverage data, surface temperature data and soil texture data into the noise prediction module and denoising module of the trained diffusion model to obtain predicted soil moisture residual data.

[0048] In the embodiment of the present application, the target time step T and the standard Gaussian noise of the target time step T are , soil moisture data of the second spatial resolution, time information and latitude information, vegetation coverage data, surface temperature data and soil texture data are input into the noise prediction module of the trained diffusion model to obtain the predicted noise of the target time step T. The predicted noise of the target time step T and the standard Gaussian noise of the target time step T are converted into Input to the denoising module to obtain the denoising result of time step T-1. Take time step T-1, the denoising result of time step T-1, soil moisture data of the second spatial resolution, time information and latitude information, vegetation coverage data, surface temperature data and soil texture data as the input of the noise prediction module, and output the predicted noise of time step T-1. Knowing the predicted noise of time step T-1, the denoising result of time step T-2 can be obtained. Traverse t from T to 1 until the predicted noise at time step t=1 is obtained. Knowing the predicted noise at time step t=1, the predicted soil moisture residual data can be obtained.

[0049] S60: Obtain downscaled soil moisture data according to the predicted soil moisture residual data and the soil moisture data at the second spatial resolution.

[0050] In an embodiment of the present application, the soil moisture data at the second spatial resolution is added to the predicted soil moisture residual data to obtain the downscaled soil moisture data.

[0051] Applying the embodiment of the present application, the embodiment of the present application first interpolates the soil moisture data of low spatial resolution to obtain the interpolated soil moisture data of high spatial resolution. By performing forward diffusion and reverse denoising processes on the soil moisture residual data between the interpolated soil moisture data of high spatial resolution and the real soil moisture data of high spatial resolution, the diffusion model is trained so that the diffusion model can predict the soil moisture residual data from the standard Gaussian noise. Based on the interpolated soil moisture data of high spatial resolution and the predicted soil moisture residual data, the downscaled soil moisture data is obtained, and the downscaling of the soil moisture data is achieved. The spatial resolution of the downscaled soil moisture data is high, which can meet the needs of refined research and application.

[0052] In one embodiment, step S20 includes step S21, which is specifically as follows:

[0053] S21: using a nearest neighbor interpolation algorithm to interpolate the soil moisture data at the first spatial resolution to obtain soil moisture data at the second spatial resolution.

[0054] Among them, the nearest neighbor interpolation algorithm means that for each pixel position of the new image, the nearest pixel point in the original image is found, and the pixel value of the point is assigned to the corresponding position of the new image.

[0055] In an embodiment of the present application, soil moisture data of a first spatial resolution is taken as an original image, and soil moisture data of a second spatial resolution is taken as a new image. The size of the new image and the corresponding relationship between the original image are determined, each pixel point of the new image is traversed, and its nearest neighbor point in the original image is calculated. The pixel value of the nearest neighbor point is assigned to the current pixel point of the new image to achieve conversion from the first spatial resolution to the second spatial resolution.

[0056] In one embodiment, before step S30, steps S31 to S32 are included, which are specifically as follows:

[0057] S31: Obtaining the longitude and latitude information of vegetation coverage data, the longitude and latitude information of surface temperature data, and the longitude and latitude information of soil texture data.

[0058] In an embodiment of the present application, the longitude and latitude information of vegetation coverage data, the longitude and latitude information of surface temperature data, and the longitude and latitude information of soil texture data can be obtained from a geographic remote sensing data sharing platform.

[0059] S32: Match the longitude and latitude information of the vegetation coverage data, the longitude and latitude information of the surface temperature data, and the longitude and latitude information of the soil texture data with the longitude and latitude information of the soil moisture data of the second spatial resolution, respectively, to obtain vegetation coverage data, surface temperature data, and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data of the second spatial resolution.

[0060] In an embodiment of the present application, by matching the longitude coordinates and the latitude coordinates, vegetation coverage data, surface temperature data and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data of the second spatial resolution are obtained.

[0061] In one embodiment, step S40 includes steps S41 to S48, which are specifically as follows:

[0062] S41: Acquire soil moisture sample data; the soil moisture sample data includes a plurality of soil moisture data at a first spatial resolution and a plurality of soil moisture data at a third spatial resolution; wherein the first spatial resolution is higher than the third spatial resolution.

[0063] In the embodiment of the present application, some soil moisture data of the first spatial resolution can be obtained from the soil moisture product SMAP, and some soil moisture data of the third spatial resolution can be obtained from the soil moisture product CCI (Climate Change Initiative). Specifically, the third spatial resolution is 25 km.

[0064] S42: interpolate the soil moisture data of the third spatial resolution to obtain soil moisture data of the fourth spatial resolution; interpolate the longitude and latitude information of the soil moisture data of the third spatial resolution to obtain the longitude and latitude information of the soil moisture data of the fourth spatial resolution; wherein the fourth spatial resolution is consistent with the first spatial resolution.

[0065] In the embodiment of the present application, the soil moisture data of the third spatial resolution is interpolated using a preset interpolation algorithm to obtain soil moisture data of the fourth spatial resolution. The longitude and latitude information of the soil moisture data of the third spatial resolution is interpolated using a preset interpolation algorithm to obtain the longitude and latitude information of the soil moisture data of the fourth spatial resolution. Specifically, the fourth spatial resolution is 9 km.

[0066] The preset interpolation algorithms include but are not limited to nearest neighbor interpolation, bilinear interpolation and bicubic interpolation algorithms.

[0067] S43: Obtaining sample soil moisture residual data according to the soil moisture data of the first spatial resolution and the soil moisture data of the fourth spatial resolution; wherein the sample soil moisture residual data is the sample soil moisture residual data of a preset time step.

[0068] In the embodiment of the present application, soil moisture data at the first spatial resolution and soil moisture data at the fourth spatial resolution are subtracted to obtain sample soil moisture residual data.

[0069] S44: Gaussian noise is gradually added to the sample soil moisture residual data through the forward diffusion module to obtain noisy soil moisture residual data at several different time steps.

[0070] In the embodiment of the present application, the noisy soil moisture residual data at different time steps can be expressed as:

[0071]

[0072] in, represents the noisy soil moisture residual data at time step t, represents the sample soil moisture residual data, , , represents the set noise addition schedule, , , T=1000, represents Gaussian noise.

[0073] S45: Acquire the time information, longitude and latitude information of the soil moisture data of the fourth spatial resolution, and sample vegetation coverage data, sample surface temperature data, and sample soil texture data having the same spatial resolution as the soil moisture data of the fourth spatial resolution.

[0074] In an embodiment of the present application, the vegetation coverage data, surface temperature data and soil texture data of 3 km are upscaled to 9 km, and based on the longitude and latitude information, sample vegetation coverage data, sample surface temperature data and sample soil texture data with the same longitude and latitude information as the soil moisture data of the fourth spatial resolution are matched from the vegetation coverage data, surface temperature data and soil texture data of 9 km to obtain sample vegetation coverage data, sample surface temperature data and sample soil texture data.

[0075] S46: Input the current time step, the noisy soil moisture residual data of the current time step, the soil moisture data of the fourth spatial resolution, the time information and longitude and latitude information of the soil moisture data of the fourth spatial resolution, the sample vegetation coverage data, the sample surface temperature data and the sample soil texture data into the noise prediction module to obtain the predicted noise; wherein the current time step is any time step sampled from several different time steps.

[0076] In the embodiment of the present application, the current time step is any time step t in 1~T. The current time step t and the noisy soil moisture residual data of the current time step t are converted into , soil moisture data of the fourth spatial resolution, time information and longitude and latitude information of the soil moisture data of the fourth spatial resolution, sample vegetation coverage data, sample surface temperature data and sample soil texture data are input into the noise prediction module, and the predicted noise is output .

[0077] S47: Calculate the loss function value according to the Gaussian noise corresponding to the current time step, the predicted noise and the preset loss function.

[0078] In the embodiment of the present application, in order to measure the difference between the predicted noise and the real noise, the mean square error loss function is used as the preset loss function. The mean square error loss function will give a greater penalty to larger deviations, prompting the network to learn more accurately. Specifically, the calculation formula of the loss function value is as follows:

[0079]

[0080] Where E represents the expected value of the random variable. represents the Gaussian noise at the current time step, represents the prediction noise.

[0081] S48: Iteratively train the noise prediction module according to the loss function value until a preset training condition is met, thereby obtaining a trained noise prediction module.

[0082] Among them, the preset training condition is that the loss function value is less than a preset threshold.

[0083] In an embodiment of the present application, the network parameters of the noise prediction module are adjusted according to the loss function value until the loss function value is less than a preset threshold, thereby obtaining a trained noise prediction module.

[0084] By training the noise prediction module through the current time step, the noisy soil moisture residual data of the current time step, the soil moisture data of the fourth spatial resolution, the time information and the longitude and latitude information of the soil moisture data of the fourth spatial resolution, the sample vegetation coverage data, the sample surface temperature data and the sample soil texture data, the trained noise prediction module can be automatically and quickly obtained.

[0085] In one embodiment, step S43 includes steps S431 to S434, which are specifically as follows:

[0086] S431: Obtain the longitude and latitude information of soil moisture data of the first spatial resolution.

[0087] In the embodiment of the present application, the longitude coordinates and the latitude coordinates of the soil moisture data of the first spatial resolution are extracted from the soil moisture product SMAP.

[0088] S432: Extracting target soil moisture data having the same longitude and latitude information as the soil moisture data of the first spatial resolution from the soil moisture data of the fourth spatial resolution according to the longitude and latitude information of the soil moisture data of the first spatial resolution.

[0089] In an embodiment of the present application, target soil moisture data having the same longitude coordinates and latitude coordinates as the soil moisture data of the first spatial resolution is matched from the soil moisture data of the fourth spatial resolution, so that the target soil moisture data and the soil moisture data of the first spatial resolution are soil moisture data of the same area.

[0090] S433: Scan the soil moisture data of the first spatial resolution and the target soil moisture data respectively according to the preset scanning window to obtain a plurality of soil moisture data blocks of the first spatial resolution and a plurality of target soil moisture data blocks.

[0091] The preset scanning window can be manually set according to actual needs. For example, the preset scanning window is 64*64.

[0092] In the embodiment of the present application, the soil moisture data of the first spatial resolution and the target soil moisture data are scanned according to a preset window to obtain a plurality of soil moisture data blocks of the first spatial resolution and a plurality of target soil moisture data blocks. For example, if the size of the soil moisture data of the first spatial resolution is 256*256 and the preset scanning window is 64*64, the number of soil moisture data blocks of the first spatial resolution is 16.

[0093] S434: Calculate the residual between each soil moisture data block of the first spatial resolution and the corresponding target soil moisture data block to obtain sample soil moisture residual data.

[0094] In the embodiment of the present application, for each soil moisture data block of the first spatial resolution, a target soil moisture data block with the same longitude coordinate and latitude coordinate is obtained. The soil moisture data of each soil moisture data block of the first spatial resolution is subtracted from the soil moisture data of the corresponding target soil moisture data block to obtain sample soil moisture residual data.

[0095] By matching the longitude and latitude information and scanning the preset scanning window, the sample soil moisture residual data can be automatically obtained.

[0096] In one embodiment, step S50 includes steps S51 to S55, which are specifically as follows:

[0097] S51: Input the target time step, standard Gaussian noise of the target time step, soil moisture data of the second spatial resolution, time information and longitude and latitude information, vegetation coverage data, surface temperature data and soil texture data into the noise prediction module to obtain the predicted noise of the target time step.

[0098] In an embodiment of the present application, the target time step, standard Gaussian noise of the target time step, soil moisture data of the second spatial resolution, time information and longitude and latitude information, vegetation coverage data, surface temperature data and soil texture data are respectively vectorized, and the corresponding vectors are input into the noise prediction module to obtain the predicted noise of the target time step.

[0099] S52: Input the predicted noise of the target time step and the standard Gaussian noise of the target time step into a denoising module to obtain a denoising result of the time step before the target time step.

[0100] In the embodiment of the present application, the denoising module denoises the standard Gaussian noise of the target time step according to the predicted noise of the target time step using a preset denoising formula to obtain the denoising result of the time step before the target time step. The preset denoising formula is as follows:

[0101]

[0102] in, represents the denoising result at time step t-1, represents the noisy soil moisture residual data at time step t, z represents the noise sampled from the normal distribution, .

[0103] Let t = T, and use the above preset denoising formula to get the denoising result of the time step before the target time step.

[0104] S53: The time step before the target time step, the denoising result of the time step before the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data are input into the noise prediction module to obtain the predicted noise of the time step before the target time step.

[0105] In an embodiment of the present application, the time step before the target time step, the denoising result of the time step before the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data are respectively vectorized, and the corresponding vectors are input into the noise prediction module to obtain the predicted noise of the time step before the target time step.

[0106] S54: inputting the predicted noise of the previous time step of the target time step and the denoising result of the previous time step of the target time step into a denoising module until a denoising result of a preset time step is obtained.

[0107] In the embodiment of the present application, the preset time step is t=1, and the denoising result of the preset time step can be obtained by repeating steps S51 to S53 using the preset denoising formula.

[0108] S55: Using the denoising result of the preset time step as the predicted soil moisture residual data.

[0109] In the embodiment of the present application, predicted soil moisture residual data can be obtained by gradually denoising the standard Gaussian noise of the target time step.

[0110] In one embodiment, step S51 includes steps S511 to S514, which are specifically as follows:

[0111] S511: Vectorize the target time step to obtain a first vector.

[0112] In the embodiment of the present application, a preset vector encoding algorithm is used to convert the target time step into a first vector, wherein the preset vector encoding algorithm includes but is not limited to a bag-of-words model and word embedding.

[0113] S512: vectorizing the standard Gaussian noise of the target time step and the soil moisture data of the second spatial resolution respectively to obtain a second vector and a third vector; and splicing the second vector and the third vector to obtain a spliced ​​vector.

[0114] In the embodiment of the present application, a preset vector encoding algorithm is used to convert the standard Gaussian noise of the target time step into a second vector, and to convert the soil moisture data of the second spatial resolution into a third vector. The preset vector encoding algorithm includes but is not limited to a bag-of-words model and word embedding.

[0115] S513: vectorize the time information of the soil moisture data of the second spatial resolution, the latitude and longitude information of the soil moisture data of the second spatial resolution, the vegetation coverage data, the surface temperature data and the soil texture data respectively to obtain a fourth vector, a fifth vector, a sixth vector, a seventh vector and an eighth vector.

[0116] Among them, the space-time encoder is an automatic encoder used to compress the input data into a low-dimensional vector.

[0117] In an embodiment of the present application, the time information of the soil moisture data of the second spatial resolution, the latitude and longitude information of the soil moisture data of the second spatial resolution, the vegetation coverage data, the surface temperature data and the soil texture data are respectively encoded by a space-time encoder to obtain a fourth vector, a fifth vector, a sixth vector, a seventh vector and an eighth vector.

[0118] S514: Input the first vector, the concatenated vector, the fourth vector, the fifth vector, the sixth vector, the seventh vector and the eighth vector into a noise prediction module to obtain predicted noise at a target time step.

[0119] In the embodiment of the present application, the first vector, the concatenated vector, the fourth vector, the fifth vector, the sixth vector, the seventh vector and the eighth vector are respectively used as input quantities of a noise prediction module, and the noise prediction module outputs predicted noise.

[0120] The embodiment of the present application can automatically determine the input amount of the noise prediction module and automatically and quickly obtain the predicted noise by vectorizing the target time step, standard Gaussian noise of the target time step, soil moisture data of the second spatial resolution, time information and longitude and latitude information, vegetation coverage data, surface temperature data and soil texture data.

[0121] The following is an embodiment of the device of the present application, which can be used to execute the content of the method in the embodiment of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the content of the method in the embodiment of the present application.

[0122] See also Figure 2 , which shows a schematic diagram of the structure of the soil moisture data downscaling device provided in an embodiment of the present application. The soil moisture data downscaling device 7 provided in an embodiment of the present application includes:

[0123] A soil moisture data acquisition module 71 is used to acquire soil moisture data of a first spatial resolution, longitude and latitude information and time information of the soil moisture data of the first spatial resolution, and standard Gaussian noise of a target time step;

[0124] The soil moisture data interpolation module 72 is used to interpolate the soil moisture data of the first spatial resolution to obtain the soil moisture data of the second spatial resolution; interpolate the longitude and latitude information of the soil moisture data of the first spatial resolution to obtain the longitude and latitude information of the soil moisture data of the second spatial resolution; the second spatial resolution is higher than the first spatial resolution;

[0125] An auxiliary data acquisition module 73, used to acquire vegetation coverage data, surface temperature data and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data of the second spatial resolution;

[0126] The diffusion model training module 74 is used to establish and train the diffusion model to obtain a trained diffusion model; wherein the diffusion model includes a forward diffusion module and a reverse diffusion module, and the reverse diffusion module includes a noise prediction module and a denoising module; Gaussian noise is gradually added to the sample soil moisture residual data through the forward diffusion module to obtain the noisy soil moisture residual data at different time steps; the noise prediction module is used to perform noise prediction on the noisy soil moisture residual data at different time steps to obtain the predicted noise; the noise prediction module is iteratively trained through the added Gaussian noise and the predicted noise to obtain the trained diffusion model;

[0127] The soil moisture residual data prediction module 75 is used to input the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module and the denoising module of the trained diffusion model to obtain the predicted soil moisture residual data;

[0128] The soil moisture data downscaling module 76 is used to obtain downscaled soil moisture data according to the predicted soil moisture residual data and the soil moisture data at the second spatial resolution.

[0129] It should be noted that the soil moisture downscaling device provided in the above embodiment only uses the division of the above functional modules as an example when executing the soil moisture downscaling method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the soil moisture downscaling device provided in the above embodiment and the soil moisture downscaling method belong to the same concept, and the implementation process thereof is detailed in the method embodiment, which will not be repeated here.

[0130] The following is an embodiment of the device of the present application, which can be used to execute the content of the method in the embodiment of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the content of the method in the embodiment of the present application.

[0131] See also Figure 3 The present application also provides an electronic device 300, which may be a computer, a mobile phone, a tablet computer, etc. In an exemplary embodiment of the present application, the electronic device 300 is a computer, which may include: at least one processor 301, at least one memory 302, at least one display, at least one network interface 303, a user interface 304, and at least one communication bus 305.

[0132] The user interface 304 is mainly used to provide an input interface for the user and obtain data input by the user. Optionally, the user interface may also include a standard wired interface or a wireless interface.

[0133] The network interface 303 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0134] The communication bus 305 is used to realize the connection and communication between these components.

[0135] Among them, the processor 301 may include one or more processing cores. The processor uses various interfaces and lines to connect various parts of the entire electronic device, and executes various functions of the electronic device and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory. Optionally, the processor can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processor (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed by the display layer; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor, but implemented separately through a chip.

[0136] Among them, the memory 302 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory can be used to store instructions, programs, codes, code sets or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc.; the data storage area may store data involved in the above-mentioned method embodiments, etc. The memory may also be optionally at least one storage device located away from the aforementioned processor. As Figure 3 As shown, the memory as a computer storage medium may include an operating system, a network communication module, a user interface module, and an operating application program.

[0137] The processor can be used to call the application of the soil moisture downscaling method stored in the memory, and specifically execute the method steps of the above-mentioned embodiment. The specific execution process can refer to the specific description shown in the embodiment, which will not be repeated here.

[0138] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0139] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A soil moisture data downscaling method, characterized in that: The steps include: Acquire soil moisture data of a first spatial resolution, latitude and longitude information and time information of the soil moisture data of the first spatial resolution, and standard Gaussian noise of a target time step; Performing interpolation processing on the soil moisture data of the first spatial resolution to obtain soil moisture data of the second spatial resolution; performing interpolation processing on the longitude and latitude information of the soil moisture data of the first spatial resolution to obtain the longitude and latitude information of the soil moisture data of the second spatial resolution; The second spatial resolution is higher than the first spatial resolution; Acquire vegetation coverage data, surface temperature data, and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data of the second spatial resolution; Establish and train a diffusion model to obtain a trained diffusion model; wherein the diffusion model includes a forward diffusion module and a reverse diffusion module, and the reverse diffusion module includes a noise prediction module and a denoising module; Gaussian noise is gradually added to the sample soil moisture residual data through the forward diffusion module to obtain the noisy soil moisture residual data at different time steps; the noise prediction module is used to perform noise prediction on the noisy soil moisture residual data at different time steps to obtain the predicted noise; the noise prediction module is iteratively trained through the added Gaussian noise and the predicted noise to obtain the trained diffusion model; Input the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module and the denoising module of the trained diffusion model to obtain predicted soil moisture residual data; Downscaled soil moisture data are obtained according to the predicted soil moisture residual data and the soil moisture data at the second spatial resolution.

2. The soil moisture data downscaling method according to claim 1, characterized in that: Before the step of acquiring vegetation coverage data, surface temperature data and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data of the second spatial resolution, the method includes: Obtaining the longitude and latitude information of vegetation coverage data, the longitude and latitude information of surface temperature data, and the longitude and latitude information of soil texture data; The longitude and latitude information of the vegetation coverage data, the longitude and latitude information of the surface temperature data, and the longitude and latitude information of the soil texture data are respectively matched with the longitude and latitude information of the soil moisture data of the second spatial resolution to obtain vegetation coverage data, surface temperature data, and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data of the second spatial resolution.

3. The soil moisture data downscaling method according to claim 1, characterized in that: The step of interpolating the soil moisture data of the first spatial resolution to obtain soil moisture data of the second spatial resolution includes: The soil moisture data at the first spatial resolution is interpolated using a nearest neighbor interpolation algorithm to obtain soil moisture data at a second spatial resolution.

4. The soil moisture data downscaling method according to claim 1, characterized in that: The step of inputting the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module and the denoising module of the trained diffusion model to obtain the predicted soil moisture residual data includes: Inputting the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module to obtain the predicted noise of the target time step; Inputting the predicted noise of the target time step and the standard Gaussian noise of the target time step into the denoising module to obtain a denoising result of a time step before the target time step; Inputting the previous time step of the target time step, the denoising result of the previous time step of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module to obtain the predicted noise of the previous time step of the target time step; Inputting the predicted noise of the previous time step of the target time step and the denoising result of the previous time step of the target time step into the denoising module until the denoising result of the preset time step is obtained; The denoising result of the preset time step is used as the predicted soil moisture residual data.

5. The soil moisture data downscaling method according to claim 4, characterized in that: The step of inputting the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module to obtain the predicted noise of the target time step includes: Vectorizing the target time step to obtain a first vector; Vectorizing the standard Gaussian noise of the target time step and the soil moisture data of the second spatial resolution respectively to obtain a second vector and a third vector; splicing the second vector and the third vector to obtain a spliced ​​vector; Vectorizing the time information of the soil moisture data of the second spatial resolution, the latitude and longitude information of the soil moisture data of the second spatial resolution, the vegetation coverage data, the surface temperature data, and the soil texture data, respectively, to obtain a fourth vector, a fifth vector, a sixth vector, a seventh vector, and an eighth vector; The first vector, the concatenated vector, the fourth vector, the fifth vector, the sixth vector, the seventh vector and the eighth vector are input into the noise prediction module to obtain the predicted noise of the target time step.

6. The soil moisture data downscaling method according to claim 1, characterized in that: The step of establishing and training a diffusion model to obtain a trained diffusion model includes: Acquire soil moisture sample data; the soil moisture sample data includes a plurality of soil moisture data of a first spatial resolution and a plurality of soil moisture data of a third spatial resolution; wherein the first spatial resolution is higher than the third spatial resolution; Performing interpolation processing on the soil moisture data of the third spatial resolution to obtain soil moisture data of a fourth spatial resolution; performing interpolation processing on the longitude and latitude information of the soil moisture data of the third spatial resolution to obtain the longitude and latitude information of the soil moisture data of the fourth spatial resolution; wherein the fourth spatial resolution is consistent with the first spatial resolution; Obtaining sample soil moisture residual data according to the soil moisture data of the first spatial resolution and the soil moisture data of the fourth spatial resolution; wherein the sample soil moisture residual data is the sample soil moisture residual data of a preset time step; The forward diffusion module is used to gradually add Gaussian noise to the sample soil moisture residual data to obtain a number of noisy soil moisture residual data at different time steps; Acquire time information, longitude and latitude information of the soil moisture data at the fourth spatial resolution, and sample vegetation coverage data, sample surface temperature data, and sample soil texture data having the same spatial resolution as the soil moisture data at the fourth spatial resolution; Inputting the current time step, the noisy soil moisture residual data of the current time step, the soil moisture data of the fourth spatial resolution, the time information and longitude and latitude information of the soil moisture data of the fourth spatial resolution, the sample vegetation coverage data, the sample surface temperature data and the sample soil texture data into a noise prediction module to obtain predicted noise; wherein the current time step is any time step sampled from the several different time steps; Calculate the loss function value according to the Gaussian noise corresponding to the current time step, the predicted noise and a preset loss function; The noise prediction module is iteratively trained according to the loss function value until a preset training condition is met, thereby obtaining a trained noise prediction module.

7. The soil moisture data downscaling method according to claim 6, characterized in that: The step of obtaining sample soil moisture residual data according to the soil moisture data of the first spatial resolution and the soil moisture data of the fourth spatial resolution comprises: Acquire the latitude and longitude information of the soil moisture data of the first spatial resolution; extracting target soil moisture data having the same longitude and latitude information as the soil moisture data of the first spatial resolution from the soil moisture data of the fourth spatial resolution according to the longitude and latitude information of the soil moisture data of the first spatial resolution; According to a preset scanning window, the soil moisture data of the first spatial resolution and the target soil moisture data are scanned respectively to obtain a plurality of soil moisture data blocks of the first spatial resolution and a plurality of target soil moisture data blocks; The residual between each soil moisture data block of the first spatial resolution and the corresponding target soil moisture data block is calculated to obtain sample soil moisture residual data.

8. A soil moisture data downscaling device, characterized in that: include: A soil moisture data acquisition module, used to acquire soil moisture data of a first spatial resolution, longitude and latitude information and time information of the soil moisture data of the first spatial resolution, and standard Gaussian noise of a target time step; A soil moisture data interpolation module, configured to interpolate the soil moisture data of the first spatial resolution to obtain soil moisture data of the second spatial resolution; interpolate the longitude and latitude information of the soil moisture data of the first spatial resolution to obtain the longitude and latitude information of the soil moisture data of the second spatial resolution; The second spatial resolution is higher than the first spatial resolution; an auxiliary data acquisition module, used to acquire vegetation coverage data, surface temperature data and soil texture data whose longitude and latitude information are the same as the longitude and latitude information of the soil moisture data of the second spatial resolution; A diffusion model training module is used to establish and train a diffusion model to obtain a trained diffusion model; wherein the diffusion model includes a forward diffusion module and a reverse diffusion module, and the reverse diffusion module includes a noise prediction module and a denoising module; Gaussian noise is gradually added to the sample soil moisture residual data through the forward diffusion module to obtain noisy soil moisture residual data at different time steps; noise is predicted on the noisy soil moisture residual data at different time steps through the noise prediction module to obtain predicted noise; the noise prediction module is iteratively trained through the added Gaussian noise and the predicted noise to obtain the trained diffusion model; A soil moisture residual data prediction module, used for inputting the target time step, the standard Gaussian noise of the target time step, the soil moisture data of the second spatial resolution, the time information and the longitude and latitude information, the vegetation coverage data, the surface temperature data and the soil texture data into the noise prediction module and the denoising module of the trained diffusion model to obtain predicted soil moisture residual data; The soil moisture data downscaling module is used to obtain downscaled soil moisture data according to the predicted soil moisture residual data and the soil moisture data of the second spatial resolution.

9. An electronic device, characterized in that: include: A processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the steps of the soil moisture data downscaling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the soil moisture data downscaling method according to any one of claims 1 to 7 are implemented.

Images