A method and device for reconstructing high-temporal and spatial resolution NDVI time series data of wetlands

Abstract

The application provides a kind of wetland high spatio-temporal resolution NDVI time series data reconstruction method and device, it is related to remote sensing image big data and wetland ecological monitoring technical field.Method includes: based on the NDVI product of study area establishes time envelope line, simultaneously using optical image and SAR polarization data, obtain submergence proportion in combination with water body scattering model;Through linear spectral mixture model separates pure vegetation NDVI;Construct multi-branch modulation neural network, with submergence proportion as modulator to carry out affine transformation to optical characteristics, input time envelope line, backscattering coefficient and pure vegetation NDVI, high spatio-temporal resolution pure vegetation NDVI sequence is reconstructed after filtering interpolation;With submergence proportion is synthesized and physical constraint is applied, the reconstruction result obtained can accurately capture the NDVI sharp drop caused by instantaneous submergence.The application can realize high-precision, continuous high spatio-temporal resolution NDVI time series data reconstruction.

Description

Technical Field

[0001] This invention relates to the fields of remote sensing image big data and wetland ecological monitoring technology, and in particular to a method and apparatus for reconstructing wetland high spatiotemporal resolution NDVI time-series data. Background Technology

[0002] Wetland ecosystems exhibit high spatiotemporal heterogeneity, with dramatic water level fluctuations and alternating vegetation growth and inundation periods, leading to nonlinear changes in vegetation indices over short periods. To accurately characterize wetland vegetation phenology and ecosystem evolution, reliable NDVI time-series data relies on both fine spatial resolution and high-density temporal series observations. However, existing remote sensing NDVI data products present contradictions; a single optical sensor cannot simultaneously capture both, and data loss is easily caused by cloud cover. For example, while medium-to-high resolution imagery such as Landsat and Sentinel-2 offers rich spatial detail (10-30m), it suffers from long revisit periods and is susceptible to cloud cover interference; MODIS data boasts a high temporal frequency (daily / 8 days) but low spatial resolution (250m-1km), making it difficult to depict fragmented wetland patches. Therefore, it is necessary to comprehensively utilize multi-source remote sensing data to reconstruct high spatiotemporal resolution NDVI time-series data. Currently, mainstream spatiotemporal fusion algorithms or technologies fall into three categories: spatiotemporal fusion algorithms for land surface reflectance such as STARFM, ESTARFM, and FSDAF; temporal smoothing and missing-filling algorithms (Savitzky-Golay filtering, harmonic analysis / HANTS); and deep learning-based fusion network algorithms. However, traditional spatiotemporal fusion techniques typically assume that land cover types do not undergo abrupt changes during the fusion period. When dealing with areas like wetlands where water levels fluctuate frequently and hydrological conditions change significantly, they often neglect the nonlinear influence of hydrophysical processes such as water level changes on NDVI values, leading to significant errors in the reconstruction results. There is an urgent need for a high spatiotemporal resolution NDVI time-series data reconstruction method that is unaffected by hydrological conditions. Summary of the Invention

[0003] To address the technical problems of existing technologies, such as abrupt changes in the water-vegetation surface caused by wetland flooding / recovery, interpolation across abrupt intervals generating results that violate physical laws, the NDVI of water-vegetation mixed pixels being a nonlinear mixed signal, the concentrated optical data loss during the most drastic periods of surface change due to torrential rains accompanied by widespread cloud cover, and the failure of observations during critical periods, and the inability of existing reconstruction results to output pixel-by-pixel uncertainty assessment results, this invention provides a method and apparatus for high spatiotemporal resolution NDVI time-series data reconstruction of wetlands. The technical solution is as follows: On the one hand, a method for reconstructing wetland high spatiotemporal resolution NDVI time-series data is provided. This method is implemented by a wetland high spatiotemporal resolution NDVI time-series data reconstruction device, and includes: S1: Establish a time envelope based on the NDVI product of the study area to obtain a time envelope with high time frequency, and acquire optical images of the study area. At the same time, obtain backscattering coefficient sequence and SAR polarization characteristic data that are synchronized or nearly synchronized with the optical images by acquiring synthetic aperture radar data. S2: Based on SAR polarization feature data and combined with water scattering model, a surface water inversion model framework is constructed by establishing it pixel by pixel or time by time, and sub-pixel calculation is performed under the surface water inversion model framework to obtain the inundation ratio. S3: Using optical images of the study area, extract the near-infrared band reflectance and the red band reflectance, calculate the NDVI value of each pixel to extract the NDVI of clear sky pixels, and store the missing values ​​in the form of NaN to obtain the NDVI dataset. Then, traverse each pixel in the NDVI dataset to obtain the NDVI time series curve of each pixel and the time index corresponding to the NDVI time series curve. S4: Based on the linear spectral mixing model, the NDVI time series curve and the time index are represented as a linear weighted sum of vegetation components and water components, and the vegetation proportion is calculated. The vegetation proportion is compared with a reliability threshold. When the vegetation proportion is not greater than the reliability threshold, the mixed pixels are filtered. When the vegetation proportion is greater than the reliability threshold, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model. S5: Using a model loss function that includes hydrological consistency constraints and temporal smoothing terms, a basic multi-branch modulation neural network containing optical coarse resolution branches, optical high resolution branches, and SAR independent branches is trained to obtain a multi-branch modulation neural network; the flooding ratio is used as the modulator input to the feature modulation layer or gating mechanism to perform affine transformation on macroscopic phenological features, spatial details, and textures to obtain an optical branch feature map after affine transformation. S6: Input the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network, extract the optical branch feature map after affine transformation, and use Savitzky-Golay filtering or cubic spline interpolation algorithm to fill in the missing vegetation components on the time axis to perform high spatiotemporal resolution NDVI time series reconstruction, obtaining a pure vegetation high spatiotemporal resolution NDVI time series dataset; then use the pure vegetation high spatiotemporal resolution NDVI time series dataset and the corresponding submergence ratio to recombine and synthesize to obtain a preliminary reconstruction result; apply physical constraints to the preliminary reconstruction result according to the actual submergence state to obtain the final reconstruction result, which includes the accurately captured NDVI drop caused by instantaneous submergence.

[0004] Preferably, in step S1, the NDVI product based on the study area is used to establish a time envelope, resulting in a time envelope with high time frequency. Optical images of the study area are acquired, and simultaneously, by acquiring synthetic aperture radar data, backscattering coefficient sequences and SAR polarization characteristic data synchronized or nearly synchronized with the optical images are obtained, including: S11: Using NDVI products from the study area, a time envelope is established to obtain a time envelope with high time frequency. The NDVI products include MODIS or VIIRS NDVI products, and the time envelope exhibits a mixed pixel effect. S12: Acquire optical images of the study area, including Landsat 8 / 9 or Sentinel-2 optical images; S13: By acquiring synthetic aperture radar data, backscattering coefficient sequence and SAR polarization feature data are obtained. The synthetic aperture radar data includes Sentinel-1 synthetic aperture radar data. The backscattering coefficient sequence is synchronized or nearly synchronized with the optical image of the study area. The SAR polarization feature data includes VV polarization features or VH polarization features.

[0005] Preferably, S2, based on SAR polarization feature data and combined with a water scattering model, constructs a surface water inversion model framework using a pixel-by-pixel or time-by-time approach, and performs sub-pixel calculations within this framework to obtain the inundation percentage, including: S21: Based on SAR polarization feature data and combined with water scattering models, a surface water inversion model framework is constructed by establishing it pixel-by-pixel or time-by-time. S22: Under the framework of the surface water inversion model, the sub-pixel surface water coverage ratio is calculated to obtain the inundation ratio.

[0006] Preferably, in step S3, optical images of the study area are used to extract near-infrared and red light reflectance. NDVI values ​​are calculated for each pixel to extract the NDVI of clear-sky pixels, and missing values ​​are stored as NaN to obtain an NDVI dataset. Then, each pixel in the NDVI dataset is traversed to obtain the NDVI time series curve for each pixel and the corresponding time index, including: S31: Using optical images of the study area, extract the near-infrared band reflectance and red band reflectance of the optical images; S32: Based on the near-infrared band reflectance and the red band reflectance, the NDVI value of each pixel in the optical image is calculated using the NDVI calculation formula to extract the NDVI of clear sky pixels, and the missing values ​​are stored in the form of NaN to obtain the NDVI dataset. S33: Traverse each cell in the NDVI dataset to obtain the NDVI time series curve of each cell and the time index corresponding to the NDVI time series curve.

[0007] Preferably, in step S4, based on a linear spectral mixing model, the NDVI time series curve and the time index are represented as a linear weighted sum of vegetation and water components, and the vegetation proportion is calculated. The vegetation proportion is compared with a reliability threshold. When the vegetation proportion is not greater than the reliability threshold, mixed pixels are filtered. When the vegetation proportion is greater than the reliability threshold, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model, including: S41: According to the linear spectral mixing model, the NDVI time series curve and the time index are represented as a linear weighted sum of vegetation components and water components, and the vegetation proportion is calculated according to the linear spectral mixing model; the linear spectral mixing model includes the product of vegetation proportion and pure vegetation NDVI value, and the product of water proportion and pure water NDVI value, wherein the pure water NDVI value includes the mean or local minimum of pure water in the study area; S42: The vegetation proportion is compared with the reliability threshold τ. When the vegetation proportion is not greater than the reliability threshold τ, the mixed pixels are filtered to eliminate the calculation noise caused by weak signal. When the vegetation proportion is greater than the reliability threshold τ, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model. The reliability threshold τ can be adjusted according to the signal-to-noise ratio.

[0008] Preferably, step S5 employs a model loss function that includes hydrological consistency constraints and temporal smoothing terms to train a basic multi-branch modulation neural network containing optical coarse-resolution, optical high-resolution, and SAR-independent branches, thereby obtaining a multi-branch modulation neural network; the inundation ratio is used as the modulator input to the feature modulation layer or gating mechanism to perform affine transformations on macroscopic phenological features, spatial details, and textures, resulting in an optical branch feature map after affine transformation, including: S51: A multi-branch modulation neural network is trained using a model loss function that includes hydrological consistency constraints and temporal smoothing terms. The multi-branch modulation neural network includes an optical coarse resolution branch, an optical high resolution branch, and a SAR-independent branch. The optical coarse resolution branch uses 1D-CNN or Transformer to extract macroscopic phenological features. The optical high resolution branch uses a U-Net structure encoder to extract spatial details and textures. The SAR-independent branch processes the backscattering coefficient sequence and the flooding ratio. The multi-branch modulation neural network also includes a feature modulation layer or gating mechanism and Monte Carlo uncertainty assessment. S52: The flooding ratio is used as a modulator and input to the feature modulation layer or gating mechanism to perform affine transformation on macroscopic phenological features, spatial details and textures to obtain an optical branch feature map after affine transformation; the feature modulation layer or gating mechanism uses the flooding ratio as a constraint condition to automatically suppress the vegetation signal weight in the optical features when displaying a high flooding probability, thereby enhancing the expression of water scattering features.

[0009] Preferably, in step S6, the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components are input into a multi-branch modulation neural network to extract optical branch feature maps after affine transformation. This is combined with Savitzky-Golay filtering or cubic spline interpolation algorithms to fill in missing vegetation components on the time axis, performing high spatiotemporal resolution NDVI time-series reconstruction to obtain a pure vegetation high spatiotemporal resolution NDVI time-series dataset. Then, the pure vegetation high spatiotemporal resolution NDVI time-series dataset and the corresponding inundation percentage are recombined and synthesized to obtain a preliminary reconstruction result. Finally, the preliminary reconstruction result is physically constrained according to the actual inundation state to obtain the final reconstruction result. The final reconstruction result includes accurately captured NDVI drops caused by instantaneous inundation, including: S61: Input the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network, extract the optical branch feature map after affine transformation, and perform high spatiotemporal resolution NDVI time series reconstruction to obtain a high spatiotemporal resolution NDVI time series dataset of pure vegetation; the reconstruction includes using Savitzky-Golay filtering or cubic spline interpolation algorithm, combined with high-resolution clear sky observation and coarse-resolution spatiotemporal support, to fill in the missing vegetation components on the time axis, and obtain the reconstructed pure vegetation NDVI data. The reconstructed pure vegetation NDVI data is a fine spatial and high temporal resolution continuous pure vegetation NDVI time series. The reconstruction also includes a basic statistical level, which includes MODIS trend, SG filtering, and missing data completion. S62: Using the reconstructed pure vegetation NDVI data and the corresponding submerged percentage at the time, the data are recombined and synthesized to obtain preliminary reconstruction results; S63: Collect the actual flooding state, apply physical constraints to the preliminary reconstruction results based on the actual flooding state, and obtain the final reconstruction results. The physical constraints are performed during the water-vegetation transition period and replace simple temporal interpolation. The final reconstruction results include the NDVI drop caused by instantaneous flooding that is accurately captured.

[0010] On the other hand, a wetland high spatiotemporal resolution NDVI time-series data reconstruction apparatus is provided. This apparatus is applied to the wetland high spatiotemporal resolution NDVI time-series data reconstruction method, and the apparatus includes: Data acquisition module: used to establish time envelope based on NDVI products of the study area, obtain time envelope with high time frequency, and acquire optical images of the study area. At the same time, by acquiring synthetic aperture radar data, it obtains backscattering coefficient sequence and SAR polarization characteristic data that are synchronized or nearly synchronized with the optical images. Inundation Ratio Module: Based on SAR polarization feature data and combined with water scattering models, it constructs a surface water inversion model framework by establishing it pixel by pixel or time by time, and performs sub-pixel calculations under the surface water inversion model framework to obtain the inundation ratio. The time series curve module is used to extract near-infrared and red reflectance from optical images of the study area, calculate NDVI values ​​for each pixel to extract the NDVI of clear sky pixels, and store missing values ​​in NaN format to obtain an NDVI dataset. The module then iterates through each pixel in the NDVI dataset to obtain the NDVI time series curve for each pixel and the corresponding time index. The NDVI module for pure vegetation components is used to represent the NDVI time series curve and the time index as a linear weighted sum of vegetation components and water components according to the linear spectral mixing model, calculate the vegetation proportion, compare the vegetation proportion with a reliability threshold, filter mixed pixels when the vegetation proportion is not greater than the reliability threshold, and calculate the NDVI value of pure vegetation components using the linear spectral mixing model when the vegetation proportion is greater than the reliability threshold. Affine Transformation Module: This module is used to train a basic multi-branch modulation neural network, which includes an optical coarse resolution branch, an optical high resolution branch, and a SAR independent branch, using a model loss function that includes hydrological consistency constraints and temporal smoothing terms. The resulting multi-branch modulation neural network is obtained by using the flooding ratio as the modulator input to the feature modulation layer or gating mechanism to perform affine transformations on macroscopic phenological features, spatial details, and textures, thereby obtaining an optical branch feature map after affine transformation. The reconstruction module is used to input the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network, extract optical branch feature maps after affine transformation, and use Savitzky-Golay filtering or cubic spline interpolation algorithms to fill in missing vegetation components on the time axis, performing high spatiotemporal resolution NDVI time-series reconstruction to obtain a pure vegetation high spatiotemporal resolution NDVI time-series dataset; then, the pure vegetation high spatiotemporal resolution NDVI time-series dataset and the corresponding submergence ratio are recombined and synthesized to obtain a preliminary reconstruction result; the preliminary reconstruction result is physically constrained according to the actual submergence state to obtain the final reconstruction result, which includes the accurately captured NDVI drop caused by instantaneous submergence.

[0011] On the other hand, a wetland high spatiotemporal resolution NDVI time series data reconstruction device is provided, the wetland high spatiotemporal resolution NDVI time series data reconstruction device includes: a processor; a memory, the memory storing computer-readable instructions, the computer-readable instructions being executed by the processor to implement the method described in any one of the above wetland high spatiotemporal resolution NDVI time series data reconstruction methods.

[0012] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: The reconstruction accuracy is unaffected by hydrological conditions, significantly eliminating errors in traditional interpolation algorithms in the water-land transition zone. Utilizing SAR inversion of inundation signals, it achieves complete, continuous temporal reconstruction unaffected by cloud cover. It boasts strong physical interpretability, clearly separating the NDVI signals of water bodies and vegetation. It outputs pixel-by-pixel uncertainty, supporting downstream data quality screening and control. It is suitable for monitoring wetland flooding cycles and long-term vegetation productivity under changing hydrological conditions (such as the impact of water conservancy projects). It effectively resolves the contradiction between "smooth vegetation growth" and "suddenness of inundation events" in wetland reconstruction. It can output a high-resolution 30-meter, 8-day wetland dataset, not only filling the gaps left by cloud cover but, more importantly, correcting the underestimation of NDVI in inundated pixels through hydrological constraints, providing more realistic biophysical parameters for wetland ecosystem monitoring.

[0013] Terminology Explanation: NDVI: Normalized Difference Vegetation Index SAR: Synthetic Aperture Radar Savitzky-Golay filter MODIS: Moderate Resolution Imaging Spectroradiometer VIIRS NDVI: Visible Infrared Imaging Radiometer Suite NDVI NaN: Not a Number 1D-CNN or Transformer: One-dimensional convolutional neural network or transformer model Landsat and Sentinel-2: Landsat and Sentinel-2 satellites STARFM: Spatial and Temporal Adaptive Reflectance Fusion Model ESTARFM: Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model FSDAF: Flexible Spatiotemporal Data Fusion HANTS: Harmonic Analysis of Time Series Algorithm Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of a method for reconstructing wetland high spatiotemporal resolution NDVI time series data according to an embodiment of the present invention; Figure 2 This invention provides a preferred neural network implementation architecture for a high spatiotemporal resolution NDVI reconstruction method. Figure 3 This is a block diagram of a wetland high spatiotemporal resolution NDVI time series data reconstruction device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a wetland high spatiotemporal resolution NDVI time series data reconstruction device provided in an embodiment of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0017] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0018] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0019] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0020] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0021] This invention provides a method for reconstructing high spatiotemporal resolution NDVI time-series data of wetlands. This method can be implemented using a high spatiotemporal resolution NDVI time-series data reconstruction device, which can be a terminal or a server. Figure 1 The flowchart shown is for a method to reconstruct wetland high spatiotemporal resolution NDVI time-series data. The processing flow of this method may include the following steps: Preferably, a time envelope is established based on the NDVI products of the study area to obtain a time envelope with high time frequency, and optical images of the study area are acquired. Simultaneously, by acquiring synthetic aperture radar data, backscattering coefficient sequences and SAR polarization characteristic data synchronized or nearly synchronized with the optical images are obtained, including: A time envelope was established using NDVI products from the study area to obtain a time envelope with high time frequency. The NDVI products include MODIS or VIIRS NDVI products, and the time envelope exhibits a mixed pixel effect. Acquire optical images of the study area, including Landsat 8 / 9 or Sentinel-2 optical images; By acquiring synthetic aperture radar (SAR) data, backscattering coefficient sequences and SAR polarization feature data are obtained. The SAR data includes Sentinel-1 SAR data, and the backscattering coefficient sequences are synchronized or nearly synchronized with the optical images of the study area. The SAR polarization feature data includes VV polarization features or VH polarization features.

[0022] In some embodiments, MODIS or VIIRS NDVI products of the study area are acquired to establish a time envelope with high temporal frequency but mixed pixel effects. Fine-resolution low-frequency imagery: Landsat 8 / 9 or Sentinel-2 optical images are acquired, and NDVI of clear-sky pixels is extracted as spatial basis data support. SAR time-series data: Sentinel-1 and other synthetic aperture radar data are acquired, and their cloud-penetrating properties are utilized to obtain backscattering coefficient sequences synchronized or nearly synchronized with the optical images.

[0023] Preferably, based on SAR polarization feature data and combined with a water scattering model, a surface water inversion model framework is constructed using a pixel-by-pixel or time-by-time approach. Sub-pixel calculations are then performed within this surface water inversion model framework to obtain the inundation percentage, including: Based on SAR polarization feature data and combined with water scattering models, a surface water inversion model framework is constructed by establishing it pixel-by-pixel or time-by-time. Within the framework of the surface water inversion model, the sub-pixel surface water coverage ratio is calculated to obtain the inundation ratio.

[0024] In some embodiments, Sentinel-1 SAR time-series data is not affected by cloud cover. Using SAR polarization characteristics (such as VV, VH) and water scattering models, a pixel-by-pixel, time-by-time surface water inversion model is established, and then the sub-pixel surface water coverage ratio is calculated. This signal, representing the percentage of submerged areas, serves as an independent observation source independent of weather conditions, characterizing the changes in the area composition of "water-vegetation" within mixed pixels.

[0025] Preferably, using optical images of the study area, near-infrared and red-light reflectance are extracted. NDVI values ​​are calculated for each pixel to extract the NDVI of clear-sky pixels, and missing values ​​are stored as NaN to obtain an NDVI dataset. Then, each pixel in the NDVI dataset is traversed to obtain the NDVI time series curve for each pixel and the corresponding time index, including: Using optical images of the study area, the near-infrared and red light reflectance of the optical images are extracted; Based on the reflectivity of the near-infrared band and the reflectivity of the red band, the NDVI value of each pixel in the optical image is calculated using the NDVI calculation formula to extract the NDVI of clear sky pixels, and missing values ​​are stored in the form of NaN to obtain the NDVI dataset. The NDVI dataset is traversed to obtain the NDVI time series curve for each pixel and the corresponding time index.

[0026] In some embodiments, optical image data primarily uses the near-infrared and red bands to calculate NDVI time-series data (some remote sensing image data provides NDVI data products). The formula for calculating NDVI is as follows:

[0027] In the above formula, For near-infrared reflectivity, Let be the reflectance in the red light band. Based on the above formula, the NDVI dataset, which is accurate but contains a large number of missing values, can be calculated. Then, pixel-by-pixel traversal is performed to construct the NDVI time series curve for each pixel, along with its corresponding time index. Missing values ​​are stored in NaN format.

[0028] Preferably, according to the linear spectral mixing model, the NDVI time series curve and the time index are represented as a linear weighted sum of vegetation and water components, and the vegetation proportion is calculated. The vegetation proportion is compared with a reliability threshold. When the vegetation proportion is not greater than the reliability threshold, mixed pixels are filtered out. When the vegetation proportion is greater than the reliability threshold, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model, including: According to the linear spectral mixing model, the NDVI time series curve and the time index are represented as a linear weighted sum of vegetation components and water components. The vegetation proportion is calculated according to the linear spectral mixing model. The linear spectral mixing model includes the product of vegetation proportion and pure vegetation NDVI value, and the product of water proportion and pure water NDVI value. The pure water NDVI value includes the mean or local minimum of pure water in the study area. The vegetation percentage is compared with the reliability threshold τ. When the vegetation percentage is not greater than the reliability threshold τ, the mixed pixels are filtered to eliminate the calculation noise caused by weak signals. When the vegetation percentage is greater than the reliability threshold τ, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model. The reliability threshold τ can be adjusted according to the signal-to-noise ratio.

[0029] It should be noted that the observed mixed pixel NDVI ( This is represented as a linear weighted average of vegetation components and water components:

[0030] In the above formula, and These represent the NDVI values ​​of pure vegetation and water components in the mixed pixels, respectively. Take the average or local minimum value of pure water in the study area.

[0031] A preset reliability threshold τ (e.g., 0.4, adjustable according to signal-to-noise ratio) is set only when the vegetation coverage rate is ( When τ > 0, the NDVI of the pure vegetation component is calculated using the linear spectral mixing model above. This filtering mechanism can eliminate calculation noise caused by weak signals during the flooding period.

[0032] Preferably, a model loss function containing hydrological consistency constraints and temporal smoothing terms is used to train a basic multi-branch modulation neural network, which includes an optical coarse-resolution branch, an optical high-resolution branch, and a SAR-independent branch, to obtain a multi-branch modulation neural network. The inundation ratio is used as the modulator input to the feature modulation layer or gating mechanism to perform affine transformations on macroscopic phenological features, spatial details, and textures, resulting in an affine-transformed optical branch feature map, including: A multi-branch modulation neural network is trained using a model loss function that includes hydrological consistency constraints and temporal smoothing terms. The multi-branch modulation neural network includes an optical coarse resolution branch, an optical high resolution branch, and a SAR-independent branch. The optical coarse resolution branch uses 1D-CNN or Transformer to extract macroscopic phenological features, the optical high resolution branch uses a U-Net structure encoder to extract spatial details and textures, and the SAR-independent branch processes the backscattering coefficient sequence and the flooding ratio. The multi-branch modulation neural network also includes a feature modulation layer or gating mechanism and Monte Carlo uncertainty assessment. The flooding percentage is used as a modulator and input to a feature modulation layer or gating mechanism to perform affine transformation on macroscopic phenological features, spatial details and textures to obtain an optical branch feature map after affine transformation. The feature modulation layer or gating mechanism uses the flooding percentage as a constraint condition to automatically suppress the vegetation signal weight in the optical features when the flooding probability is high, thereby enhancing the expression of water scattering features.

[0033] Preferably, the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components are input into a multi-branch modulation neural network to extract optical branch feature maps after affine transformation. The missing vegetation components on the time axis are then filled in using Savitzky-Golay filtering or cubic spline interpolation algorithms to perform high spatiotemporal resolution NDVI time-series reconstruction, resulting in a pure vegetation high spatiotemporal resolution NDVI time-series dataset. The pure vegetation high spatiotemporal resolution NDVI time-series dataset and the corresponding inundation percentage are then recombined and synthesized to obtain a preliminary reconstruction result. The preliminary reconstruction result is then physically constrained according to the actual inundation state to obtain the final reconstruction result. The final reconstruction result includes accurately captured NDVI drops caused by instantaneous inundation, including: The temporal envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components are input into a multi-branch modulation neural network to extract optical branch feature maps after affine transformation. High spatiotemporal resolution NDVI time series reconstruction is then performed to obtain a high spatiotemporal resolution NDVI time series dataset of pure vegetation. The reconstruction includes using Savitzky-Golay filtering or cubic spline interpolation algorithms, combined with high-resolution clear-sky observations and coarse-resolution temporal spatial support, to fill in missing vegetation components on the time axis, resulting in reconstructed pure vegetation NDVI data. The reconstructed pure vegetation NDVI data is a fine-spatial, high-temporal resolution continuous pure vegetation NDVI time series. The reconstruction also includes a basic statistical level, which includes MODIS trend analysis, SG filtering, and missing data completion. The reconstructed pure vegetation NDVI data and the corresponding submerged percentage at the corresponding time were recombined and synthesized to obtain preliminary reconstruction results; The actual flooding state is collected, and the preliminary reconstruction results are physically constrained based on the actual flooding state to obtain the final reconstruction results. The physical constraints are carried out during the water-vegetation transition period and replace simple temporal interpolation. The final reconstruction results include the NDVI drop caused by instantaneous flooding that is accurately captured.

[0034] In some embodiments, high-resolution clear-sky observations and coarse-resolution temporal spatial support are combined to reconstruct a fine-spatial, high-temporal-resolution continuous pure vegetation NDVI time series. Because After removing transient flooding fluctuations, the numerical evolution of vegetation data is primarily controlled by plant physiological patterns, exhibiting strong temporal smoothness and physical predictability. Using Savitzky-Golay filtering or cubic spline interpolation algorithms, missing vegetation components are filled in on the time axis to generate reconstructed pure vegetation NDVI data. .

[0035] To improve reconstruction efficiency and cross-regional generalization ability, a multi-branch deep learning modulation neural network is selected to construct a pure vegetation high spatiotemporal resolution NDVI time series dataset (optical coarse resolution branch, optical high resolution branch, and SAR independent branch). Specifically, 1D-CNN or Transformer is used to process the MODIS time series to extract macroscopic phenological features; a U-Net structure encoder is used to extract spatial details and textures of Sentinel / Landsat; and a dedicated processing method for SAR backscattering and derived inundation frequency sequences is used to provide all-weather physical background constraints. A Feature Modulation Layer (FiLM) gating mechanism is introduced with the inundation ratio as a constraint condition, and the inundation features output by the SAR branch are used as a "modulator" to perform affine transformation on the feature map extracted by the optical branch. When the SAR signal shows a high inundation probability, the modulation layer automatically suppresses the vegetation signal weight in the optical features and enhances the expression of water scattering features. At the same time, a hydrological consistency constraint term is added to the model loss function. ):

[0036] in, This is the total loss value. For mean square error loss, For hydrological constraint weighting coefficients, For smoothness weighting coefficients, As a smoothing constraint, it penalizes the physical deviation between the reconstructed NDVI and the theoretical range calculated from the SAR inundation percentage. For example, if the SAR shows total inundation, but the model predicts an NDVI significantly higher than the water body threshold, a high penalty is applied. The temporal smoothing term constrains the gradient changes of pure vegetation components during the latency period, ensuring the continuity of the physiological cycle.

[0037] It should be noted that the reconstructed pure vegetation component The flooding percentage extracted by SAR at the corresponding time. Recombinant synthesis was carried out.

[0038]

[0039] During water-vegetation transition periods (such as rising or receding water levels), the reconstruction results are physically constrained by the actual inundation state rather than simply determined by temporal interpolation, thus accurately capturing the sudden drop in NDVI caused by instantaneous inundation.

[0040] It should be further explained that, taking into account SAR inversion error, optical pixel cloud cover mask and model residual, an uncertainty estimate is output for each grid point (every 30 meters for 8 days) to quantify the credibility of the reconstruction results.

[0041] The above is an introduction to the method embodiments. The following describes the solution described in this application through device embodiments.

[0042] Figure 3 This is a block diagram of a wetland high spatiotemporal resolution NDVI time-series data reconstruction apparatus according to an exemplary embodiment. The apparatus is used in a wetland high spatiotemporal resolution NDVI time-series data reconstruction method. (Refer to...) Figure 3 The device includes a data acquisition module 310, a flooding percentage module 320, a time series curve module 330, an NDVI module for pure vegetation components 340, an affine transformation module 350, and a reconstruction module 360.

[0043] Data acquisition module 310: used to establish a time envelope based on NDVI products of the study area, obtain a time envelope with high time frequency, acquire optical images of the study area, and at the same time, obtain backscattering coefficient sequence and SAR polarization characteristic data that are synchronized or nearly synchronized with the optical images by acquiring synthetic aperture radar data. Inundation percentage module 320: Based on SAR polarization feature data and combined with water scattering model, it constructs a surface water inversion model framework by establishing it pixel by pixel or time by time, and performs sub-pixel calculation under the surface water inversion model framework to obtain the inundation percentage. Time series curve module 330: Used to extract near-infrared band reflectance and red band reflectance from optical images of the study area, calculate NDVI value for each pixel to extract clear sky pixel NDVI and store missing values ​​in NaN form to obtain NDVI dataset, and traverse each pixel in the NDVI dataset to obtain NDVI time series curve for each pixel and time index corresponding to the NDVI time series curve; The NDVI module 340 for pure vegetation components is used to represent the NDVI time series curve and the time index as a linear weighted sum of vegetation components and water components according to the linear spectral mixing model and calculate the vegetation proportion. The vegetation proportion is compared with a reliability threshold. When the vegetation proportion is not greater than the reliability threshold, the mixed pixels are filtered. When the vegetation proportion is greater than the reliability threshold, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model. Affine Transformation Module 350: Used to train a basic multi-branch modulation neural network containing optical coarse resolution branch, optical high resolution branch and SAR independent branch using a model loss function that includes hydrological consistency constraint term and time series smoothing term, to obtain a multi-branch modulation neural network; the flooding ratio is used as the modulator input to the feature modulation layer or gating mechanism to perform affine transformation on macroscopic phenological features, spatial details and texture to obtain an optical branch feature map after affine transformation. Reconstruction Module 360: This module inputs the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network. It extracts optical branch feature maps after affine transformation and uses Savitzky-Golay filtering or cubic spline interpolation algorithms to fill in missing vegetation components on the time axis, performing high spatiotemporal resolution NDVI time-series reconstruction to obtain a pure vegetation high spatiotemporal resolution NDVI time-series dataset. Then, it uses the pure vegetation high spatiotemporal resolution NDVI time-series dataset and the corresponding inundation percentage to recombine and synthesize, obtaining a preliminary reconstruction result. Based on the actual inundation state, it applies physical constraints to the preliminary reconstruction result to obtain the final reconstruction result, which includes accurately captured NDVI drops caused by instantaneous inundation.

[0044] A wetland high spatiotemporal resolution NDVI time series data reconstruction device, the wetland high spatiotemporal resolution NDVI time series data reconstruction device comprising: a processor; a memory, the memory storing computer-readable instructions, the computer-readable instructions being executed by the processor to implement the method described in any one of the above wetland high spatiotemporal resolution NDVI time series data reconstruction methods.

[0045] Figure 4 This is a schematic diagram of the structure of a wetland high spatiotemporal resolution NDVI time-series data reconstruction device provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the wetland high spatiotemporal resolution NDVI time-series data reconstruction equipment may include the above-mentioned Figure 3 The illustrated wetland high spatiotemporal resolution NDVI time series data reconstruction apparatus may optionally include a first processor 2001.

[0046] Optionally, the wetland high spatiotemporal resolution NDVI time series data reconstruction device 410 may also include a memory 2002 and a transceiver 2003.

[0047] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0048] The following is combined Figure 4 A detailed description of each component of the wetland high spatiotemporal resolution NDVI time-series data reconstruction device 410 is provided below: The first processor 2001 is the control center of the wetland high spatiotemporal resolution NDVI time-series data reconstruction device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0049] Optionally, the first processor 2001 can perform various functions of the wetland high spatiotemporal resolution NDVI time series data reconstruction device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0050] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 4 CPU0 and CPU1 are shown in the diagram.

[0051] In a specific implementation, as one example, the wetland high spatiotemporal resolution NDVI time-series data reconstruction device 410 may also include multiple processors, for example... Figure 4 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0052] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0053] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected to the interface circuit of the wetland high spatiotemporal resolution NDVI timing data reconstruction device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0054] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0055] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 4 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0056] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently, and its interface circuit can be used with the wetland high spatiotemporal resolution NDVI timing data reconstruction device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0057] It should be noted that, Figure 4 The structure of the wetland high spatiotemporal resolution NDVI time series data reconstruction device 410 shown in the figure does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0058] Furthermore, the technical effect of the wetland high spatiotemporal resolution NDVI time series data reconstruction device 410 can be referred to the technical effect of the wetland high spatiotemporal resolution NDVI time series data reconstruction method described in the above method embodiments, and will not be repeated here.

[0059] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0060] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0061] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0062] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0063] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0064] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0065] 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 technical solution. 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.

[0066] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0067] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of 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 device, 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.

[0068] 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.

[0069] In addition, 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.

[0070] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for reconstructing wetland high spatiotemporal resolution NDVI time-series data, characterized in that, The method includes: S1: Establish a time envelope based on the NDVI product of the study area to obtain a time envelope with high time frequency, and acquire optical images of the study area. At the same time, obtain backscattering coefficient sequence and SAR polarization characteristic data that are synchronized or nearly synchronized with the optical images by acquiring synthetic aperture radar data. S2: Based on SAR polarization feature data and combined with water scattering model, a surface water inversion model framework is constructed by establishing it pixel by pixel or time by time, and sub-pixel calculation is performed under the surface water inversion model framework to obtain the inundation ratio. S3: Using optical images of the study area, extract the near-infrared band reflectance and the red band reflectance, calculate the NDVI value of each pixel to extract the NDVI of clear sky pixels, obtain the NDVI dataset, and traverse each pixel in the NDVI dataset to obtain the NDVI time series curve of each pixel and the time index corresponding to the NDVI time series curve. S4: Based on the linear spectral mixing model, the NDVI time series curve and the time index are represented as a linear weighted sum of vegetation components and water components, and the vegetation proportion is calculated. The vegetation proportion is compared with a reliability threshold. When the vegetation proportion is not greater than the reliability threshold, the mixed pixels are filtered. When the vegetation proportion is greater than the reliability threshold, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model. S5: Using a model loss function that includes hydrological consistency constraints and temporal smoothing terms, a basic multi-branch modulation neural network containing optical coarse resolution branches, optical high resolution branches, and SAR independent branches is trained to obtain a multi-branch modulation neural network; the flooding ratio is used as the modulator input to the feature modulation layer or gating mechanism to perform affine transformation on macroscopic phenological features, spatial details, and textures to obtain an optical branch feature map after affine transformation. S6: Input the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network, extract the optical branch feature map after affine transformation, and use Savitzky-Golay filtering or cubic spline interpolation algorithm to fill in the missing vegetation components on the time axis to perform high spatiotemporal resolution NDVI time series reconstruction, obtaining a pure vegetation high spatiotemporal resolution NDVI time series dataset; then use the pure vegetation high spatiotemporal resolution NDVI time series dataset and the corresponding submergence ratio to recombine and synthesize to obtain a preliminary reconstruction result; apply physical constraints to the preliminary reconstruction result according to the actual submergence state to obtain the final reconstruction result, which includes the accurately captured NDVI drop caused by instantaneous submergence.

2. The wetland high spatiotemporal resolution NDVI time-series data reconstruction method according to claim 1, characterized in that, The S1 uses NDVI products based on the study area to establish a time envelope, obtaining a time envelope with high time frequency, and acquires optical images of the study area. Simultaneously, by acquiring synthetic aperture radar data, it obtains backscattering coefficient sequences and SAR polarization characteristic data that are synchronized or nearly synchronized with the optical images, including: S11: Using NDVI products from the study area, a time envelope is established to obtain a time envelope with high time frequency. The NDVI products include MODIS or VIIRS NDVI products, and the time envelope exhibits a mixed pixel effect. S12: Acquire optical images of the study area, including Landsat 8 / 9 or Sentinel-2 optical images; S13: By acquiring synthetic aperture radar data, backscattering coefficient sequence and SAR polarization feature data are obtained. The synthetic aperture radar data includes Sentinel-1 synthetic aperture radar data. The backscattering coefficient sequence is synchronized or nearly synchronized with the optical image of the study area. The SAR polarization feature data includes VV polarization features or VH polarization features.

3. The method for reconstructing wetland high spatiotemporal resolution NDVI time-series data according to claim 1, characterized in that, S2, based on SAR polarization feature data and combined with a water scattering model, constructs a surface water inversion model framework using a pixel-by-pixel or time-by-time approach. Sub-pixel calculations are then performed within this framework to obtain the inundation percentage, including: S21: Based on SAR polarization feature data and combined with water scattering models, a surface water inversion model framework is constructed by establishing it pixel-by-pixel or time-by-time. S22: Under the framework of the surface water inversion model, the sub-pixel surface water coverage ratio is calculated to obtain the inundation ratio.

4. The wetland high spatiotemporal resolution NDVI time-series data reconstruction method according to claim 1, characterized in that, The S3 method uses optical images of the study area to extract near-infrared and red reflectance, calculates NDVI values ​​for each pixel to extract clear-sky pixel NDVI, obtains an NDVI dataset, and iterates through each pixel in the NDVI dataset to obtain an NDVI time series curve for each pixel and a corresponding time index, including: S31: Using optical images of the study area, extract the near-infrared band reflectance and red band reflectance of the optical images; S32: Based on the near-infrared band reflectance and the red band reflectance, the NDVI value of each pixel in the optical image is calculated using the NDVI calculation formula to extract the NDVI of clear sky pixels, and the missing values ​​are stored in the form of NaN to obtain the NDVI dataset. S33: Traverse each cell in the NDVI dataset to obtain the NDVI time series curve of each cell and the time index corresponding to the NDVI time series curve.

5. The method for reconstructing wetland high spatiotemporal resolution NDVI time-series data according to claim 1, characterized in that, S4, based on a linear spectral mixing model, represents the NDVI time series curve and the time index as a linear weighted sum of vegetation and water components, calculates the vegetation proportion, compares the vegetation proportion with a reliability threshold, filters mixed pixels when the vegetation proportion is not greater than the reliability threshold, and calculates the NDVI value of the pure vegetation component using the linear spectral mixing model when the vegetation proportion is greater than the reliability threshold, including: S41: According to the linear spectral mixing model, the NDVI time series curve and the time index are represented as a linear weighted sum of vegetation components and water components, and the vegetation proportion is calculated according to the linear spectral mixing model; the linear spectral mixing model includes the product of vegetation proportion and pure vegetation NDVI value, and the product of water proportion and pure water NDVI value, wherein the pure water NDVI value includes the mean or local minimum of pure water in the study area; S42: The vegetation proportion is compared with the reliability threshold τ. When the vegetation proportion is not greater than the reliability threshold τ, the mixed pixels are filtered to eliminate the calculation noise caused by weak signal. When the vegetation proportion is greater than the reliability threshold τ, the NDVI value of the pure vegetation component is calculated using the linear spectral mixing model. The reliability threshold τ can be adjusted according to the signal-to-noise ratio.

6. The method for reconstructing wetland high spatiotemporal resolution NDVI time-series data according to claim 1, characterized in that, The S5 method employs a model loss function that includes hydrological consistency constraints and temporal smoothing terms to train a basic multi-branch modulation neural network containing optical coarse-resolution, optical high-resolution, and SAR-independent branches, thereby obtaining a multi-branch modulation neural network. The inundation ratio is used as the modulator input to the feature modulation layer or gating mechanism to perform affine transformations on macroscopic phenological features, spatial details, and textures, resulting in an affine-transformed optical branch feature map, including: S51: A multi-branch modulation neural network is trained using a model loss function that includes hydrological consistency constraints and temporal smoothing terms. The multi-branch modulation neural network includes an optical coarse resolution branch, an optical high resolution branch, and a SAR-independent branch. The optical coarse resolution branch uses 1D-CNN or Transformer to extract macroscopic phenological features. The optical high resolution branch uses a U-Net structure encoder to extract spatial details and textures. The SAR-independent branch processes the backscattering coefficient sequence and the flooding ratio. The multi-branch modulation neural network also includes a feature modulation layer or gating mechanism and Monte Carlo uncertainty assessment. S52: The flooding ratio is used as a modulator and input to the feature modulation layer or gating mechanism to perform affine transformation on macroscopic phenological features, spatial details and textures to obtain an optical branch feature map after affine transformation; the feature modulation layer or gating mechanism uses the flooding ratio as a constraint condition to automatically suppress the vegetation signal weight in the optical features when displaying a high flooding probability, thereby enhancing the expression of water scattering features.

7. The method for reconstructing wetland high spatiotemporal resolution NDVI time-series data according to claim 1, characterized in that, S6 inputs the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network, extracts optical branch feature maps after affine transformation, and uses Savitzky-Golay filtering or cubic spline interpolation algorithms to fill in missing vegetation components on the time axis, performing high spatiotemporal resolution NDVI time-series reconstruction to obtain a pure vegetation high spatiotemporal resolution NDVI time-series dataset; then, it uses the pure vegetation high spatiotemporal resolution NDVI time-series dataset and the corresponding inundation percentage to recombine and synthesize, obtaining a preliminary reconstruction result; and finally, it applies physical constraints to the preliminary reconstruction result based on the actual inundation state to obtain the final reconstruction result, which includes accurately captured NDVI drops caused by instantaneous inundation, including: S61: Input the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network, extract the optical branch feature map after affine transformation, and perform high spatiotemporal resolution NDVI time series reconstruction to obtain a high spatiotemporal resolution NDVI time series dataset of pure vegetation; the reconstruction includes using Savitzky-Golay filtering or cubic spline interpolation algorithm, combined with high-resolution clear sky observation and coarse-resolution spatiotemporal support, to fill in the missing vegetation components on the time axis, and obtain the reconstructed pure vegetation NDVI data. The reconstructed pure vegetation NDVI data is a fine spatial and high temporal resolution continuous pure vegetation NDVI time series. The reconstruction also includes a basic statistical level, which includes MODIS trend, SG filtering, and missing data completion. S62: Using the reconstructed pure vegetation NDVI data and the corresponding submerged percentage at the time, the data are recombined and synthesized to obtain preliminary reconstruction results; S63: Collect the actual flooding state, apply physical constraints to the preliminary reconstruction results based on the actual flooding state, and obtain the final reconstruction results. The physical constraints are performed during the water-vegetation transition period and replace simple temporal interpolation. The final reconstruction results include the NDVI drop caused by instantaneous flooding that is accurately captured.

8. A wetland high spatiotemporal resolution NDVI time-series data reconstruction apparatus, wherein the wetland high spatiotemporal resolution NDVI time-series data reconstruction apparatus is used to implement the wetland high spatiotemporal resolution NDVI time-series data reconstruction method as described in any one of claims 1-7, characterized in that, The device includes: Data acquisition module: used to establish time envelope based on NDVI products of the study area, obtain time envelope with high time frequency, and acquire optical images of the study area. At the same time, by acquiring synthetic aperture radar data, it obtains backscattering coefficient sequence and SAR polarization characteristic data that are synchronized or nearly synchronized with the optical images. Inundation Ratio Module: Based on SAR polarization feature data and combined with water scattering models, it constructs a surface water inversion model framework by establishing it pixel by pixel or time by time, and performs sub-pixel calculations under the surface water inversion model framework to obtain the inundation ratio. The time series curve module is used to extract near-infrared and red reflectance from optical images of the study area, calculate NDVI values ​​for each pixel to extract the NDVI of clear sky pixels, and store missing values ​​in NaN format to obtain an NDVI dataset. The module then iterates through each pixel in the NDVI dataset to obtain the NDVI time series curve for each pixel and the corresponding time index. The NDVI module for pure vegetation components is used to represent the NDVI time series curve and the time index as a linear weighted sum of vegetation components and water components according to the linear spectral mixing model, calculate the vegetation proportion, compare the vegetation proportion with a reliability threshold, filter mixed pixels when the vegetation proportion is not greater than the reliability threshold, and calculate the NDVI value of pure vegetation components using the linear spectral mixing model when the vegetation proportion is greater than the reliability threshold. Affine Transformation Module: This module is used to train a basic multi-branch modulation neural network, which includes an optical coarse resolution branch, an optical high resolution branch, and a SAR independent branch, using a model loss function that includes hydrological consistency constraints and temporal smoothing terms. The resulting multi-branch modulation neural network is obtained by using the flooding ratio as the modulator input to the feature modulation layer or gating mechanism to perform affine transformations on macroscopic phenological features, spatial details, and textures, thereby obtaining an optical branch feature map after affine transformation. The reconstruction module is used to input the time envelope, backscattering coefficient sequence, and NDVI values ​​of pure vegetation components into a multi-branch modulation neural network, extract optical branch feature maps after affine transformation, and use Savitzky-Golay filtering or cubic spline interpolation algorithms to fill in missing vegetation components on the time axis, performing high spatiotemporal resolution NDVI time-series reconstruction to obtain a pure vegetation high spatiotemporal resolution NDVI time-series dataset; then, the pure vegetation high spatiotemporal resolution NDVI time-series dataset and the corresponding submergence ratio are recombined and synthesized to obtain a preliminary reconstruction result; the preliminary reconstruction result is physically constrained according to the actual submergence state to obtain the final reconstruction result, which includes the accurately captured NDVI drop caused by instantaneous submergence.

9. A wetland high spatiotemporal resolution NDVI time series data reconstruction device, characterized in that, The wetland high spatiotemporal resolution NDVI time-series data reconstruction device includes: a processor; a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

Images