Soil moisture reconstruction methods, apparatus, computer equipment, and storage media

Abstract

This invention relates to the field of geographic information technology, and particularly to a method, apparatus, computer equipment, and primary storage medium for soil moisture reconstruction. The method includes: acquiring multimodal data collected by multi-band satellite sensors; inputting first-resolution band data and second-resolution band data into a preset band resolution reconstruction model for resolution reconstruction to obtain reconstructed resolution band data; resampling soil attribute data and topographic data to the resolution of the reconstructed resolution band data to obtain resampled soil attribute data and resampled topographic data; and inputting the reconstructed resolution band data, resampled soil attribute data, resampled topographic data, geographic location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model for data reconstruction to obtain reconstructed soil moisture data, thereby meeting the high-frequency acquisition needs of high-resolution soil moisture data in multiple fields.

Description

Technical Field

[0001] This invention relates to the field of geographic information technology, and in particular to a method, apparatus, computer equipment, and storage medium for soil moisture reconstruction. Background Technology

[0002] Soil moisture is a physical quantity that indicates the dryness or wetness of soil. It is the life source for terrestrial plants. For terrestrial plants, including food crops, precipitation and irrigation must be converted into soil moisture before they can be absorbed, and applied fertilizers must also be absorbed in the form of soil solution. Therefore, soil moisture is often used as a classic indicator to describe the degree of drought and plays an indispensable role in agricultural drought and flood monitoring and early warning systems.

[0003] Satellite microwave remote sensing is a crucial method for acquiring large-scale, long-term soil moisture data products. However, daily-scale soil moisture data products retrieved from single satellite signals generally suffer from gaps and low spatial resolution (approximately 25 km), making it difficult to achieve seamless, high-resolution daily surface coverage. To improve spatial coverage, soil moisture retrieval data based on the fusion of multi-source satellite microwave data and medium-resolution imaging spectrometer data has emerged. By matching and fusing multiple reliable data sources, high-spatial-coverage daily-scale soil moisture fusion products have been developed, providing important data references for exploring the spatiotemporal evolution of surface water cycle patterns. However, the spatial resolution of most satellite microwave signal fusion-based soil moisture retrieval data remains at the kilometer or even hundreds of meter grid scale, which is insufficient to meet the needs of regional-scale drought and flood monitoring. Summary of the Invention

[0004] Based on this, the purpose of this invention is to provide a soil moisture reconstruction method, apparatus, computer equipment, and storage medium. Based on a band resolution reconstruction model, it achieves super-resolution downscaling of band data and integrates multivariate spatiotemporal information under the guidance of geoscience knowledge. Combined with the constructed soil moisture reconstruction model, it efficiently captures the nonlinear coupling relationship of the spatiotemporal dynamic texture of soil moisture, thereby improving the accuracy of high-resolution reconstruction of soil moisture data and meeting the high-frequency acquisition needs of high-resolution soil moisture data in multiple fields.

[0005] In a first aspect, embodiments of this application provide a method for soil moisture reconstruction, comprising the following steps:

[0006] The system acquires multimodal data collected by multi-band satellite sensors, including first-resolution band data, second-resolution band data, soil moisture data to be reconstructed, soil property data, topographic data, and geographic location data.

[0007] The first resolution band data and the second resolution band data are input into a preset band resolution reconstruction model for resolution reconstruction to obtain reconstructed resolution band data. The band resolution reconstruction model is a noise prediction model built based on the U-net network structure, using the first resolution band data and the second resolution band data as training data.

[0008] The soil property data and topographic data are resampled to the resolution of the reconstructed resolution band data to obtain resampled soil property data and resampled topographic data.

[0009] The reconstructed resolution band data, resampled soil property data, resampled topographic data, geographic location data, and soil moisture data to be reconstructed are input into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data.

[0010] Secondly, embodiments of this application provide a soil moisture reconstruction device, comprising:

[0011] The data acquisition module is used to acquire multimodal data collected by multi-band satellite sensors, wherein the multimodal data includes first resolution band data, second resolution band data, soil moisture data to be reconstructed, soil property data, topographic data, and geographic location data;

[0012] The resolution reconstruction module is used to input the first resolution band data and the second resolution band data into a preset band resolution reconstruction model to reconstruct the resolution and obtain reconstructed resolution band data. The band resolution reconstruction model is a noise prediction model built based on the U-net network structure, using the first resolution band data and the second resolution band data as training data.

[0013] The resampling module is used to resample the soil property data and topographic data to the resolution of the reconstructed resolution band data to obtain resampled soil property data and resampled topographic data.

[0014] The soil moisture reconstruction module is used to input the reconstructed resolution band data, resampled soil attribute data, resampled topographic data, geographical location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data.

[0015] Thirdly, embodiments of this application provide a computer device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, it implements the steps of the soil moisture reconstruction method as described in the first aspect.

[0016] Fourthly, embodiments of this application provide a storage medium storing a computer program that, when executed by a processor, implements the steps of the soil moisture reconstruction method as described in the first aspect.

[0017] In this application embodiment, a soil moisture reconstruction method, apparatus, computer equipment, and storage medium are provided. Based on a band resolution reconstruction model, super-resolution downscaling of band data is achieved. Under the guidance of geoscience knowledge, multivariate spatiotemporal information is integrated. Combined with the constructed soil moisture reconstruction model, the nonlinear coupling relationship of the spatiotemporal dynamic texture of soil moisture is efficiently captured, thereby improving the accuracy of high-resolution reconstruction of soil moisture data and meeting the high-frequency acquisition needs of high-resolution soil moisture data in multiple fields.

[0018] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0019] Figure 1 A schematic flowchart of a soil moisture reconstruction method provided in one embodiment of this application;

[0020] Figure 2 A schematic flowchart of a soil moisture reconstruction method provided in another embodiment of this application;

[0021] Figure 3 This is a schematic flowchart of step S2 in a soil moisture reconstruction method provided in one embodiment of this application;

[0022] Figure 4 This is a schematic flowchart of step S21 in a soil moisture reconstruction method provided in one embodiment of this application;

[0023] Figure 5 This is a schematic flowchart of step S22 in a soil moisture reconstruction method provided in one embodiment of this application;

[0024] Figure 6 This is a schematic flowchart of step S23 in a soil moisture reconstruction method provided in one embodiment of this application;

[0025] Figure 7 This is a schematic flowchart of step S4 in a soil moisture reconstruction method provided in one embodiment of this application;

[0026] Figure 8 This is a schematic diagram of the structure of a soil moisture reconstruction device provided in one embodiment of this application;

[0027] Figure 9 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0029] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0030] It should be understood that although the terms first, sample, third, etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as sample information, and similarly, sample information may also be referred to as first information. Depending on the context, the words “if” or “suppose” as used herein may be interpreted as “when” or “in response to determination”.

[0031] Please see Figure 1 , Figure 1 The following is a schematic flowchart of a soil moisture reconstruction method provided in one embodiment of this application, the method comprising the following steps:

[0032] S1: Obtain multimodal data acquired by multi-band satellite sensors.

[0033] The subject of the soil moisture reconstruction method is the reconstruction equipment (hereinafter referred to as the reconstruction equipment). In an optional embodiment, the reconstruction equipment may be a computer device, a server, or a server cluster composed of multiple computer devices.

[0034] In this embodiment, the reconstruction device obtains multimodal data collected by multi-band satellite sensors, wherein the multimodal data includes first resolution band data, second resolution band data, soil moisture data to be reconstructed, soil property data, topographic data, and geographical location data.

[0035] Specifically, the first resolution band data uses visible-near infrared medium resolution data (data from bands 1-7 of the MCD43A4 band data), with a spatial resolution of 500m×500m and a temporal resolution of day scale.

[0036] The second resolution band data uses visible-near infrared high-resolution data (HLS, Harmonized Landsat Sentinel-2 band data and corresponding bands 1-7 of MCD43A4 band data), with a spatial resolution of 30m and a temporal resolution of 2-3 days.

[0037] The soil moisture data to be reconstructed uses SMAP or Sentinel remote sensing soil moisture data, with a spatial resolution of 3km×3km and a temporal resolution of 12 days.

[0038] The soil property data were obtained from the Harmonized World Soil Database (HWSD V2.0), with a spatial resolution of 1 km × 1 km.

[0039] The terrain data used is SRTM terrain data with a spatial resolution of 30m × 30m. The spatial resolution of the geographic location data is also 30m × 30m.

[0040] S2: Input the first resolution band data and the second resolution band data into the preset band resolution reconstruction model to perform resolution reconstruction and obtain the reconstructed resolution band data.

[0041] In this embodiment, the reconstruction device inputs the first resolution band data and the second resolution band data into a preset band resolution reconstruction model for resolution reconstruction to obtain reconstructed resolution band data. The band resolution reconstruction model is a noise prediction model constructed based on a U-net network structure, using the first resolution band data and the second resolution band data as training data. Specifically, the band resolution reconstruction model is an optimal noise prediction model constructed by using gradient descent and Bayesian parameter optimization methods to determine structural hyperparameters such as the number of first encoder / decoder, the number and size of two-dimensional convolutional kernels, as well as model training hyperparameters such as the initial learning rate and optimizer.

[0042] Please see Figure 2 , Figure 2 A schematic flowchart of a soil moisture reconstruction method provided in another embodiment of this application further includes step S5, which, prior to step S2, specifically comprises the following:

[0043] S5: Using bilinear interpolation, the first resolution band data is resampled, and spatiotemporal matching and normalization are performed based on the second resolution band data and the resampled first resolution band data to obtain normalized first resolution band data and second resolution band data.

[0044] In this embodiment, the reconstruction device uses bilinear interpolation to resample the first resolution band data, and performs spatiotemporal matching and normalization based on the second resolution band data and the resampled first resolution band data to obtain normalized first resolution band data and second resolution band data.

[0045] The band resolution reconstruction model includes a forward diffusion module, a first encoding module, and a decoding module; the first encoding module includes a first convolutional layer and several sequentially connected first encoders; the first decoder includes several sequentially connected decoders and a second convolutional layer. Please refer to [link to relevant documentation]. Figure 3 , Figure 3 The flowchart of S2 in the soil moisture reconstruction method provided in one embodiment of this application is as follows: S21 to S24 are detailed below.

[0046] S21: Input the normalized first resolution band data and the second resolution band data into the forward diffusion module for forward diffusion processing to obtain forward diffusion data.

[0047] The forward diffusion module is used to perform the forward process, i.e. the diffusion process, which refers to the process of gradually adding Gaussian noise to the original data until the data becomes random noise.

[0048] In this embodiment, the reconstruction device inputs the normalized first resolution band data and the second resolution band data into the forward diffusion module for forward diffusion processing to obtain forward diffusion data. The forward diffusion data includes band noise data corresponding to several diffusion steps to capture the complex relationship between the normalized first resolution band data and the second resolution band data, thereby improving the accuracy of band data resolution reconstruction.

[0049] Please see Figure 4 , Figure 4 The schematic diagram of step S21 in the soil moisture reconstruction method provided in one embodiment of this application includes steps S211 to S212, as follows:

[0050] S211: Construct an image pair based on the normalized first resolution band data and the second resolution band data, wherein the image pair includes a first resolution band image and a second resolution band image.

[0051] In this embodiment, the reconstruction device constructs an image pair by cropping through a sliding window based on the normalized first resolution band data and the second resolution band data, wherein the image pair includes the first resolution band image and the second resolution band image.

[0052] S212: Based on the image pair, the preset number of diffusion steps, and the forward diffusion algorithm, obtain band noise data corresponding to several diffusion steps.

[0053] In this embodiment, the reconstruction device obtains band-denoised data corresponding to several diffusion steps based on the image pair, a preset number of diffusion steps, and a forward diffusion algorithm. This achieves the effect of multi-step diffusion, ensures the stability of the band-denoised data during the diffusion process, saves system resources, and improves the efficiency of the diffusion process. The forward diffusion algorithm is as follows:

[0054]

[0055] In the formula, x t Add noise to the band corresponding to the t-th diffusion step. Let β be the Gaussian noise correlation parameter corresponding to the t-th diffusion step. i Let y be the variance corresponding to the i-th diffusion step. HR For the second resolution band image, x LR Let be the first resolution band image, and ∈ be random noise parameters that conform to a Gaussian distribution.

[0056] S22: Input the forward diffusion data into the first convolutional layer for convolution processing to obtain a first feature array. Use the first feature array as the first input data of the first first encoder for encoding processing to obtain the sub-band encoded data output by the first first encoder. Use the first feature array and the sub-band encoded data output by the first first encoder as the first input data of the next first encoder, and repeat the encoding processing to obtain a number of sub-band encoded data output by the first encoder.

[0057] In this embodiment, the reconstruction device inputs the forward diffusion data into the first convolutional layer for convolution processing to obtain a first feature array. The first feature array is used as the first input data of the first first encoder for encoding processing to obtain the sub-band encoded data output by the first first encoder. The first feature array and the sub-band encoded data output by the first first encoder are used as the first input data of the next first encoder, and the encoding processing is repeated to obtain a number of sub-band encoded data output by the first encoder.

[0058] The first encoder includes a plurality of sequentially connected first residual blocks and a downsampling layer; the first feature array includes convolutional feature data corresponding to a plurality of diffusion steps; please refer to Figure 5 , Figure 5 The flowchart of S22 in the soil moisture reconstruction method provided in one embodiment of this application is as follows: S221 to S222 are detailed below:

[0059] S221: Input the first input data into the first residual block, perform residual processing based on the first feature array, sub-band coding data and preset residual block algorithm in the first input data to obtain the intermediate feature array output by the first residual block, use the intermediate feature array output by the first residual block and the first feature array in the first input data as the third input data of the next first residual block, repeat the residual processing to obtain the intermediate feature data output by the last residual block.

[0060] The residual block algorithm is as follows:

[0061]

[0062] In the formula, L i+1 L is the intermediate feature array output by the (i+1)th residual block. i Let X be the intermediate feature array output by the residual block of the i-th first encoder, and t be the first feature array. e This is a time-encoded array, which includes time codes obtained by transforming the diffusion steps using Transformer sinusoidal position encoding. Let be the first nonlinear mapping function, representing the nonlinear mapping after the Conv2D layer and the Mish activation function. is the second nonlinear mapping function, representing the nonlinear mapping processed by the FC layer and the ReLU activation function.

[0063] In this embodiment, the reconstruction device inputs the first input data into the first residual block, performs residual processing based on the first feature array, subband coding data and preset residual block algorithm in the first input data, and obtains the intermediate feature array output by the first residual block. The intermediate feature array output by the first residual block and the first feature array in the first input data are used as the third input data of the next first residual block, and the residual processing is repeated to obtain the intermediate feature data output by the last residual block.

[0064] S222: Input the intermediate feature data output from the last residual block into the downsampling layer for downsampling to obtain the subband encoded data output by the first encoder.

[0065] In this embodiment, the reconstruction device inputs the intermediate feature data output by the last residual block into the downsampling layer for downsampling to obtain the subband coded data output by the first encoder.

[0066] S23: The subband encoded data output by the last first encoder is used as the second input data of the first decoder for decoding to obtain the subband decoded data output by the first decoder. The subband decoded data output by the first decoder and the subband encoded data output by the previous decoder are used as the second input data of the next decoder, and the decoding process is repeated to obtain the subband decoded data output by the last first encoder.

[0067] In this embodiment, the reconstruction device uses the sub-band coded data output by the last first encoder as the second input data of the first decoder for decoding to obtain the sub-band decoded data output by the first decoder. The sub-band decoded data output by the first decoder and the sub-band coded data output by the previous decoder are used as the second input data of the next decoder, and the decoding process is repeated to obtain the sub-band decoded data output by the last first encoder.

[0068] The decoder includes several sequentially connected second residual blocks and an upsampling layer; please refer to [link / reference]. Figure 6 , Figure 6 The flowchart of S23 in the soil moisture reconstruction method provided in one embodiment of this application is as follows: S231 is included.

[0069] S231: Obtain the subband decoding data output by the first encoder based on the subband decoding data, subband coding data and preset decoding algorithm in the second input data.

[0070] The decoding algorithm is as follows:

[0071] D n =Upsampling(Resblock(Resblock(D n+1 E n )))

[0072] In the formula, D n D represents the subband decoding data output by the nth decoder. n+1 E represents the subband decoding data output by the (n+1)th decoder. n This is the subband encoded data output by the nth first encoder. Upsampling(·) is the processing function of the upsampling layer, and Resblock(·) is the processing function of the second residual block.

[0073] In this embodiment, the reconstruction device, based on the subband decoding data, subband coding data, and preset decoding algorithm in the second input data, inputs the subband decoding data and subband coding data in the second input data into the first second residual block of the current decoder for residual processing, inputs the processed result into the next second residual block for residual processing, and inputs the processed result into the upsampling layer for upsampling to obtain the subband decoding data output by the current first encoder.

[0074] S24: Input the sub-band decoding data output by the last first encoder into the second convolutional layer for convolution processing, and use the output of the second convolutional layer as the reconstructed resolution band data.

[0075] In this embodiment, the reconstruction device inputs the sub-band decoding data output by the last first encoder into the second convolutional layer for convolution processing, and uses the output of the second convolutional layer as the reconstructed resolution band data.

[0076] S3: Resample the soil property data and topographic data to the resolution of the reconstructed resolution band data to obtain resampled soil property data and resampled topographic data.

[0077] In this embodiment, the reconstruction device resamples the soil property data and topographic data to the resolution of the reconstruction resolution band data to obtain resampled soil property data and resampled topographic data. Specifically, the reconstruction device combines the reconstruction resolution band data reconstructed at a daily scale and 30m super-resolution, and resamples the soil property data and topographic data to 30m based on the bilinear interpolation method to maintain the same spatial resolution as the reconstruction resolution band data, thereby obtaining resampled soil property data and resampled topographic data.

[0078] S4: Input the reconstructed resolution band data, resampled soil property data, resampled topographic data, geographical location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data.

[0079] The soil moisture reconstruction model is a high-precision fusion model constructed based on a preset deep learning transformation model network structure. It adjusts the sample structure by utilizing the model training accuracy and efficiency, and searches for the optimal solution of deep network layers, nonlinear activation functions, gradient optimization methods, etc., based on Bayesian optimization methods.

[0080] In this embodiment, the reconstruction device inputs the reconstruction resolution band data, resampled soil property data, resampled terrain data, geographical location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data.

[0081] The soil moisture reconstruction model includes a temporal feature extraction module, a spatial feature extraction module, and a data reconstruction module. The temporal feature extraction module includes an embedding layer, a second encoding module, a fully connected layer, and a regularization layer. The second encoding module includes several sequentially connected second encoders. The spatial feature extraction module includes a convolutional module, a two-dimensional global average pooling layer, a fully connected layer, and a regularization layer. The convolutional module includes several sub-convolutional units, each including a two-dimensional convolutional layer and a two-dimensional global max pooling layer. The data reconstruction module includes a fully connected layer, an activation layer, and a two-dimensional convolutional layer. (See also...) Figure 7 , Figure 7 The flowchart of S4 in the soil moisture reconstruction method provided in one embodiment of this application is as follows: S41 to S43 are detailed below.

[0082] S41: The reconstructed resolution band data is used as a time-series dynamic variable. The time-series dynamic variable, the soil moisture data to be reconstructed, and the data are input into the time feature extraction module. Spatial mapping and position encoding are performed according to the embedding layer to obtain mapped coded data. The mapped coded data is input into the second encoding module for encoding processing to obtain the coded data output by the last second encoder. The coded data output by the last second encoder is processed sequentially through the fully connected layer and the regularization layer of the time feature extraction module to obtain time feature data.

[0083] In this embodiment, the reconstruction device uses the reconstruction resolution band data as a time-series dynamic variable, and inputs the time-series dynamic variable, the soil moisture data to be reconstructed, and the time feature extraction module. Based on the embedding layer, spatial mapping and position encoding are performed to obtain mapped and encoded data, thereby obtaining high-dimensional feature data.

[0084] The reconstruction device inputs the mapped and encoded data into the second encoding module for encoding processing to obtain the encoded data output by the last second encoder. Specifically, the second encoder includes a multi-head attention sublayer and a feedforward sublayer. The reconstruction device uses the mapped and encoded data as the input data for the first second encoder of the second encoding module, processes it sequentially through the multi-head attention sublayer and the feedforward sublayer to obtain the output data of the first second encoder, and uses the output data of the first second encoder as the input data for the next second encoder, repeating the encoding process to obtain the encoded data output by the last second encoder.

[0085] The reconstruction device processes the encoded data output by the last second encoder sequentially through the fully connected layer and the regularization layer of the time feature extraction module to obtain time feature data, thereby capturing the temporal texture features between the soil moisture data to be reconstructed and the time-series dynamic variables, and improving the accuracy and reliability of soil moisture reconstruction.

[0086] S42: The resampled soil attribute data, resampled topographic data, and geographic location data are used as time-series static variables. The time-series static variables and the soil moisture data to be reconstructed are respectively input into the spatial feature extraction module. Convolution processing is performed according to several sub-convolution units of the convolution module. The convolution data output by several sub-convolution units are concatenated to obtain the first concatenated data. The first concatenated data is processed sequentially through the two-dimensional global average pooling layer, the fully connected layer, and the regularization layer to obtain spatial feature data.

[0087] In this embodiment, the reconstruction device uses the resampled soil attribute data, resampled terrain data, and geographic location data as time-series static variables. The time-series static variables and the soil moisture data to be reconstructed are respectively input into the spatial feature extraction module. Convolution processing is performed according to several sub-convolution units of the convolution module. The convolution data output by several sub-convolution units are spliced ​​together to obtain the first spliced ​​data.

[0088] The reconstruction equipment processes the first stitched data sequentially through the two-dimensional global average pooling layer, the fully connected layer, and the regularization layer to obtain spatial feature data, thereby capturing the spatial texture features between the soil moisture data to be reconstructed and the time-series static variables, and improving the accuracy and reliability of soil moisture reconstruction.

[0089] S43: Input the temporal feature data and spatial feature data into the data reconstruction module respectively. Perform fully connected processing on the temporal feature data and spatial feature data according to the fully connected layer. Concatenate the fully connected feature data corresponding to the temporal feature data and the fully connected feature data corresponding to the spatial feature data output by the fully connected layer to obtain the second concatenated data. Process the second concatenated data sequentially through the activation layer and the two-dimensional convolutional layer to obtain the reconstructed soil moisture data.

[0090] In this embodiment, the reconstruction device inputs the temporal feature data and spatial feature data into the data reconstruction module, respectively. The temporal feature data and spatial feature data are processed using the fully connected layer. The fully connected feature data corresponding to the temporal feature data and the fully connected feature data corresponding to the spatial feature data output by the fully connected layer are then spliced ​​together to obtain second spliced ​​data. The second spliced ​​data is then processed sequentially through the activation layer and the two-dimensional convolutional layer to obtain the reconstructed soil moisture data.

[0091] Super-resolution downscaling of band data is achieved based on a band resolution reconstruction model. Multivariate spatiotemporal information is integrated under the guidance of geoscience knowledge. Combined with the constructed soil moisture reconstruction model, the nonlinear coupling relationship of the spatiotemporal dynamic texture of soil moisture is efficiently captured, thereby improving the accuracy of high-resolution reconstruction of soil moisture data and meeting the high-frequency acquisition needs of high-resolution soil moisture data in multiple fields.

[0092] Please refer to Figure 8 , Figure 8 This is a schematic diagram of a soil moisture reconstruction device provided in one embodiment of this application. The device can be implemented entirely or partially through software, hardware, or a combination of both. The device 8 includes:

[0093] The data acquisition module 81 is used to acquire multimodal data collected by multi-band satellite sensors, wherein the multimodal data includes first resolution band data, second resolution band data, soil moisture data to be reconstructed, soil property data, topographic data, and geographical location data.

[0094] The resolution reconstruction module 82 is used to input the first resolution band data and the second resolution band data into a preset band resolution reconstruction model to perform resolution reconstruction and obtain reconstructed resolution band data. The band resolution reconstruction model is a noise prediction model built based on the U-net network structure using the first resolution band data and the second resolution band data as training data.

[0095] The resampling module 83 is used to resample the soil property data and topographic data to the resolution of the reconstructed resolution band data to obtain resampled soil property data and resampled topographic data.

[0096] The soil moisture reconstruction module 84 is used to input the reconstructed resolution band data, resampled soil attribute data, resampled topographic data, geographical location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data.

[0097] In this embodiment, a data acquisition module obtains multimodal data collected by multi-band satellite sensors. This multimodal data includes first-resolution band data, second-resolution band data, soil moisture data to be reconstructed, soil property data, topographic data, and geographic location data. A resolution reconstruction module inputs the first-resolution band data and the second-resolution band data into a preset band resolution reconstruction model for resolution reconstruction, obtaining reconstructed resolution band data. This band resolution reconstruction model is a noise prediction model built based on a U-net network structure, using the first-resolution band data and the second-resolution band data as training data. A resampling module resamples the soil property data and topographic data to the resolution of the reconstructed resolution band data, obtaining resampled soil property data and resampled topographic data. Finally, a soil moisture reconstruction module inputs the reconstructed resolution band data, resampled soil property data, resampled topographic data, geographic location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model for data reconstruction, obtaining reconstructed soil moisture data. It can comprehensively consider the impact of topographic data and precipitation data on soil moisture reconstruction. Based on existing topographic data, precipitation data and soil moisture reconstruction, it can establish a regression mapping model between soil moisture and surface parameters, and achieve high-precision reconstruction of soil moisture.

[0098] Please refer to Figure 9 , Figure 9 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application. The computer device 9 includes: a processor 91, a memory 92, and a computer program 93 stored in the memory 92 and executable on the processor 91; the computer device can store multiple instructions, which are adapted to be loaded and executed by the processor 91. Figures 1 to 7 The method steps of the illustrated embodiment can be found in the following documentation for detailed execution. Figures 1 to 7 The specific details of the illustrated embodiments will not be elaborated here.

[0099] The processor 91 may include one or more processing cores. The processor 91 connects to various parts of the server using various interfaces and lines, and executes various functions and processes data of the soil moisture reconstruction device 4 by running or executing instructions, programs, code sets, or instruction sets stored in the memory 92, and by calling data stored in the memory 92. Optionally, the processor 91 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 91 may integrate one or a combination of several of the following: a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required to be displayed on the touch screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 91 and may be implemented as a separate chip.

[0100] The memory 92 may include random access memory (RAM) or read-only memory. Optionally, the memory 92 may include a non-transitory computer-readable storage medium. The memory 92 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 92 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 touch instructions), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data involved in the various method embodiments described above, etc. Optionally, the memory 92 may also be at least one storage device located remotely from the aforementioned processor 91.

[0101] This application embodiment also provides a storage medium that can store multiple instructions, which are adapted to be loaded and executed by a processor as described above. Figures 1 to 7 The method steps of the illustrated embodiment can be found in the following documentation for detailed execution. Figures 1 to 7 The specific details of the illustrated embodiments will not be elaborated here.

[0102] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0103] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0104] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the algorithm. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0105] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0106] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0107] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0108] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms.

[0109] This invention is not limited to the above-described embodiments. If any modifications or variations to this invention do not depart from the spirit and scope of this invention, and if such modifications and variations fall within the scope of the claims and equivalent technologies of this invention, then this invention also intends to include such modifications and variations.

Claims

1. A method for soil moisture reconstruction, characterized in that, The following steps are involved: The system acquires multimodal data collected by multi-band satellite sensors, including first-resolution band data, second-resolution band data, soil moisture data to be reconstructed, soil property data, topographic data, and geographic location data. The first resolution band data and the second resolution band data are input into a preset band resolution reconstruction model. The band resolution reconstruction model is a noise prediction model built based on the U-net network structure, using the first resolution band data and the second resolution band data as training data. The band resolution reconstruction model includes a forward diffusion module, a first encoding module, and a decoding module. The first encoding module includes a first convolutional layer and several first encoders connected in sequence. The decoding module includes several decoders connected in sequence and a second convolutional layer. The normalized first-resolution band data and second-resolution band data are input into the forward diffusion module for forward diffusion processing to obtain forward diffusion data, wherein the forward diffusion data includes band noise data corresponding to several diffusion steps. The forward-spreading data is input into the first convolutional layer for convolution processing to obtain a first feature array. The first feature array is used as the first input data of the first first encoder for encoding processing to obtain the sub-band encoded data output by the first first encoder. The first feature array and the sub-band encoded data output by the first first encoder are used as the first input data of the next first encoder, and the encoding processing is repeated to obtain a number of sub-band encoded data output by the first encoder. The subband encoded data output by the last first encoder is used as the second input data of the first decoder for decoding to obtain the subband decoded data output by the first decoder. The subband decoded data output by the first decoder and the subband encoded data output by the previous decoder are used as the second input data of the next decoder, and the decoding process is repeated to obtain the subband decoded data output by the last first encoder. The sub-band decoding data output by the last first encoder is input into the second convolutional layer for convolution processing, and the output of the second convolutional layer is used as the reconstructed resolution band data to obtain the reconstructed resolution band data. The soil property data and topographic data are resampled to the resolution of the reconstructed resolution band data to obtain resampled soil property data and resampled topographic data. The reconstructed resolution band data, resampled soil property data, resampled topographic data, geographic location data, and soil moisture data to be reconstructed are input into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data.

2. The soil moisture reconstruction method according to claim 1, characterized in that: The first resolution band data uses visible light-near infrared medium resolution data; the second resolution band data uses visible light-near infrared high resolution data. Before inputting the first resolution band data and the second resolution band data into a preset band resolution reconstruction model for resolution reconstruction to obtain the reconstructed resolution band data, the steps include: The first resolution band data is resampled using a bilinear interpolation method. Then, spatiotemporal matching and normalization are performed based on the second resolution band data and the resampled first resolution band data to obtain normalized first resolution band data and second resolution band data.

3. The soil moisture reconstruction method according to claim 1, characterized in that, The step of inputting the normalized first-resolution band data and second-resolution band data into the forward diffusion module for forward diffusion processing to obtain forward-dividered data includes the following steps: Based on the normalized first-resolution band data and second-resolution band data, an image pair is constructed, wherein the image pair includes a first-resolution band image and a second-resolution band image. Based on the image pair, the preset number of diffusion steps, and the forward diffusion algorithm, band-added noisy data corresponding to several diffusion steps are obtained, wherein the forward diffusion algorithm is: In the formula, x t Add noise to the band corresponding to the t-th diffusion step. Let β be the Gaussian noise correlation parameter corresponding to the t-th diffusion step. i Let y be the variance corresponding to the i-th diffusion step. HR For the second resolution band image, x LR Let be the first resolution band image, and ∈ be random noise parameters that conform to a Gaussian distribution.

4. The soil moisture reconstruction method according to claim 3, characterized in that: The first encoder includes a plurality of first residual blocks connected in sequence and a downsampling layer; the first feature array includes convolutional feature data corresponding to a plurality of diffusion steps; The step of using the first feature array and the sub-band coded data output by the first encoder as the first input data for the next encoder, and repeating the encoding process to obtain a plurality of sub-band coded data output by the first encoder, includes the following steps: The first input data is input into the first residual block. Residual processing is performed based on the first feature array, sub-band encoded data, and a preset residual block algorithm in the first input data to obtain the intermediate feature array output by the first residual block. The intermediate feature array output by the first residual block and the first feature array in the first input data are used as the third input data for the next first residual block. Residual processing is repeated to obtain the intermediate feature data output by the last residual block. The residual block algorithm is as follows: In the formula, L i+1 L is the intermediate feature array output by the (i+1)th residual block. i Let X be the intermediate feature array output by the residual block of the i-th first encoder, and t be the first feature array. e This is a time-encoded array, which includes time codes obtained by transforming the diffusion steps using Transformer sinusoidal position encoding. Let be the first nonlinear mapping function, representing the nonlinear mapping after the Conv2D layer and the Mish activation function. The second nonlinear mapping function represents the nonlinear mapping processed by the FC layer and the ReLU activation function; The intermediate feature data output from the last residual block is input into the downsampling layer for downsampling to obtain the subband coded data output by the first encoder.

5. The soil moisture reconstruction method according to claim 4, characterized in that: The decoder includes several second residual blocks connected in sequence and an upsampling layer; The process of using the subband decoded data output by the first decoder and the subband encoded data output by the previous decoder as the second input data for the next decoder, and repeating the decoding process to obtain the last subband decoded data output by the first encoder, includes the following steps: Based on the subband decoded data, subband encoded data, and a preset decoding algorithm in the second input data, the subband decoded data output by the first encoder is obtained, wherein the decoding algorithm is: D n =Upsampling(Resblock(Resblock(D n+1 ,E n ))) In the formula, D n D represents the subband decoding data output by the nth decoder. n+1 E represents the subband decoding data output by the (n+1)th decoder. n This is the subband encoded data output by the nth first encoder. Upsampling(·) is the processing function of the upsampling layer, and Resblock(·) is the processing function of the second residual block.

6. The soil moisture reconstruction method according to claim 5, characterized in that: The soil moisture reconstruction model includes a temporal feature extraction module, a spatial feature extraction module, and a data reconstruction module. The temporal feature extraction module includes an embedding layer, a second encoding module, a fully connected layer, and a regularization layer. The second encoding module includes several sequentially connected second encoders. The spatial feature extraction module includes a convolutional module, a two-dimensional global average pooling layer, a fully connected layer, and a regularization layer. The convolutional module includes several sub-convolutional units, each including a two-dimensional convolutional layer and a two-dimensional global max pooling layer. The data reconstruction module includes a fully connected layer, an activation layer, and a two-dimensional convolutional layer. The process of inputting the reconstructed resolution band data, resampled soil attribute data, resampled topographic data, geographic location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data includes the following steps: The reconstructed resolution band data is used as a time-series dynamic variable. The time-series dynamic variable, the soil moisture data to be reconstructed, and the data are input into the time feature extraction module. Spatial mapping and position encoding are performed according to the embedding layer to obtain mapped and encoded data. The mapped and encoded data is input into the second encoding module for encoding processing to obtain the encoded data output by the last second encoder. The encoded data output by the last second encoder is processed sequentially through the fully connected layer and the regularization layer of the time feature extraction module to obtain time feature data. The resampled soil attribute data, resampled topographic data, and geographic location data are used as time-series static variables. The time-series static variables and the soil moisture data to be reconstructed are respectively input into the spatial feature extraction module. Convolution processing is performed according to several sub-convolutional units of the convolution module. The convolutional data output by several sub-convolutional units are concatenated to obtain the first concatenated data. The first concatenated data is then processed sequentially through the two-dimensional global average pooling layer, the fully connected layer, and the regularization layer to obtain spatial feature data. The temporal and spatial feature data are respectively input into the data reconstruction module. The temporal and spatial feature data are processed by the fully connected layer. The fully connected feature data corresponding to the temporal feature data and the fully connected feature data corresponding to the spatial feature data output by the fully connected layer are spliced ​​together to obtain the second spliced ​​data. The second spliced ​​data is then processed by the activation layer and the two-dimensional convolutional layer to obtain the reconstructed soil moisture data.

7. A soil moisture reconstruction device, characterized in that, include: The data acquisition module is used to acquire multimodal data collected by multi-band satellite sensors, wherein the multimodal data includes first resolution band data, second resolution band data, soil moisture data to be reconstructed, soil property data, topographic data, and geographic location data; The resolution reconstruction module is used to input the first resolution band data and the second resolution band data into a preset band resolution reconstruction model. The band resolution reconstruction model is a noise prediction model built based on the U-net network structure, using the first resolution band data and the second resolution band data as training data. The band resolution reconstruction model includes a forward diffusion module, a first encoding module, and a decoding module. The first encoding module includes a first convolutional layer and several first encoders connected in sequence. The decoding module includes several decoders connected in sequence and a second convolutional layer. The normalized first-resolution band data and second-resolution band data are input into the forward diffusion module for forward diffusion processing to obtain forward diffusion data, wherein the forward diffusion data includes band noise data corresponding to several diffusion steps. The forward-spreading data is input into the first convolutional layer for convolution processing to obtain a first feature array. The first feature array is used as the first input data of the first first encoder for encoding processing to obtain the sub-band encoded data output by the first first encoder. The first feature array and the sub-band encoded data output by the first first encoder are used as the first input data of the next first encoder, and the encoding processing is repeated to obtain a number of sub-band encoded data output by the first encoder. The subband encoded data output by the last first encoder is used as the second input data of the first decoder for decoding to obtain the subband decoded data output by the first decoder. The subband decoded data output by the first decoder and the subband encoded data output by the previous decoder are used as the second input data of the next decoder, and the decoding process is repeated to obtain the subband decoded data output by the last first encoder. The sub-band decoding data output by the last first encoder is input into the second convolutional layer for convolution processing, and the output of the second convolutional layer is used as the reconstructed resolution band data to obtain the reconstructed resolution band data. The resampling module is used to resample the soil property data and topographic data to the resolution of the reconstructed resolution band data to obtain resampled soil property data and resampled topographic data. The soil moisture reconstruction module is used to input the reconstructed resolution band data, resampled soil attribute data, resampled topographic data, geographical location data, and soil moisture data to be reconstructed into a preset soil moisture reconstruction model to reconstruct the data and obtain reconstructed soil moisture data.

8. A computer device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor; the computer program, when executed by the processor, implements the steps of the soil moisture reconstruction method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores a computer program that, when executed by a processor, implements the steps of the soil moisture reconstruction method as described in any one of claims 1 to 6.

Images