NDSI remote sensing data reconstruction method, device and equipment and storage medium
By using a spatiotemporal extreme gradient boosting model and leveraging NDSI-related variables and spatiotemporal auxiliary information, the problem of data loss caused by cloud occlusion in MODIS snow cover products was solved, and high-precision NDSI data reconstruction was achieved.
Patent Information
- Application Number
- CN202310099262.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-02-01
- Publication Date
- 2025-11-28
- Estimated Expiration
- 2043-02-01
AI Technical Summary
Existing MODIS snow cover products cannot reconstruct spatiotemporally continuous NDSI data when data loss is caused by cloud cover.
A spatiotemporal extreme gradient boosting model is adopted, which utilizes NDSI-related variable data and spatiotemporal auxiliary information to construct a model to recover the missing NDSI information.
It achieves spatiotemporal continuous data reconstruction under the condition of long-term continuous missing NDSI data, with high accuracy and robust recovery of NDSI data.
Smart Images
Figure CN116129285B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of remote sensing, and in particular to an NDSI remote sensing data reconstruction method, device, equipment and storage medium. BACKGROUND
[0002] Snow, as an important part of the cryosphere, plays an important role in global climate and heat budget. The high albedo and adiabatic property of snow affect the surface energy balance and global atmospheric circulation. In addition, snow is also a sensitive indicator of climate change, which directly affects geophysical and hydrological processes. Therefore, it is of great significance to accurately collect spatiotemporal continuous snow data for snow spatiotemporal variation analysis, water resource management, climate change monitoring, etc.
[0003] Satellite remote sensing is a powerful tool for continuous monitoring of snow spatiotemporal behavior. Among various snow products, the Moderate Resolution Imaging Spectroradiometer (MODIS) snow dataset is widely used in various hydrological and climatic applications due to its relatively high spatial and temporal resolution. However, studies have shown that cloud obscuration in daily MODIS snow products results in data missing up to 39% to 70%, which limits the application of MODIS snow products in hydrological simulation, climate change research, snow disaster monitoring, etc.
[0004] Currently, there have been researches on cloud removal and reconstruction of normalized snow index (NDSI) information of MODIS snow products, such as spatiotemporal adaptive reflectance fusion model, two-stage fusion framework based on Gaussian kernel function, etc. However, the existing methods may have no similar pixels or geographically related pixels in the case of long-time continuous missing of spatiotemporal data, resulting in failure to reconstruct spatiotemporally continuous data. SUMMARY
[0005] The technical problem to be solved by the present application is to reconstruct spatiotemporally continuous NDSI data in the case of long-time continuous missing of spatiotemporal data. To solve the technical problem, the technical solution adopted by the present application is to provide an NDSI remote sensing data reconstruction method, device, equipment and storage medium, which realizes spatiotemporal seamless reconstruction of NDSI remote sensing data based on a spatiotemporal extreme gradient boosting model.
[0006] According to a first aspect of the present application, the present application provides an NDSI remote sensing data reconstruction method, comprising the following steps:
[0007] Obtaining NDSI data of a region to be reconstructed, obtaining variable data related to the NDSI data as auxiliary data, and unifying the spatial resolutions of the NDSI data and the auxiliary data;
[0008] Extracting spatiotemporal auxiliary data of each pixel;
[0009] constructing a spatio-temporal extreme gradient boosting model;
[0010] training and testing the model with the NDSI data, the auxiliary data, and the spatio-temporal auxiliary data;
[0011] retrieving the missing NDSI information by training and testing the model with the auxiliary data and the spatio-temporal auxiliary data corresponding to the missing pixels.
[0012] Preferably, the auxiliary data comprises elevation data, solar zenith angle data, solar azimuth angle data, sensor zenith angle data, sensor azimuth angle data, land cover type remote sensing data, and surface albedo remote sensing data.
[0013] Further, the spatio-temporal auxiliary data comprises temporal auxiliary data and spatial auxiliary data.
[0014] Preferably, the temporal auxiliary data comprises time corresponding to each pixel and NDSI information of adjacent temporal valid pixels, and the spatial auxiliary data comprises longitude and latitude of each pixel and NDSI information of adjacent spatial valid pixels.
[0015] Further, the spatio-temporal extreme gradient boosting model has input data comprising the auxiliary data and the spatio-temporal auxiliary data, and preferably, the existing NDSI data, the auxiliary data corresponding thereto, and the spatio-temporal auxiliary data corresponding thereto are randomly divided into 70% for training the model and 30% for testing the model.
[0016] Further, the spatio-temporal extreme gradient boosting model can be formulated as:
[0017] NDSI = f(Day, Lat, Lon, DEM, LAC, SoA, SoZ, SeA, SeZ, Albedo, SN, TN)
[0018]
[0019]
[0020] wherein Day represents the day of the year, Lat and Lon represent longitude and latitude, respectively, DEM represents elevation, LAC represents land cover type, SoA, SoZ, SeA, and SeZ represent solar zenith angle and azimuth angle, sensor zenith angle and azimuth angle, respectively, Albedo represents surface albedo, (m, n) represents spatial position of a pixel, t represents serial number of a pixel at time t, num represents number of valid pixels, y m,n,tNDSI value of spatial position (m, n) at t day, SN represents NDSI information of adjacent spatial effective pixels, TN represents NDSI information of adjacent time effective pixels, and NDSI represents predicted NDSI value.
[0021] According to a second aspect of the present application, the present application provides an NDSI remote sensing data reconstruction device, comprising the following modules:
[0022] A data acquisition module is configured to acquire NDSI data of a region to be reconstructed, acquire variable data related to the NDSI data as auxiliary data, and unify the spatial resolutions of the NDSI data and the auxiliary data.
[0023] A data extraction module is configured to extract spatio-temporal auxiliary data of each pixel.
[0024] A model construction module is configured to construct a spatio-temporal extreme gradient boosting model.
[0025] A training and testing module is configured to input the NDSI data, the auxiliary data and the spatio-temporal auxiliary data to train and test the model.
[0026] A data reconstruction module is configured to input auxiliary data and spatio-temporal auxiliary data corresponding to a missing pixel element into the trained and tested model to recover the missing NDSI information.
[0027] According to a third aspect of the present application, the present application provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the NDSI remote sensing data reconstruction method when executing the program.
[0028] According to a fourth aspect of the present application, the present application provides a storage medium having a computer program stored thereon, wherein the computer program is executable on a processor to implement the steps of the NDSI remote sensing data reconstruction method.
[0029] The technical solution provided by the present application has the following beneficial effects:
[0030] The present application makes full use of variables related to NDSI, further introduces spatio-temporal auxiliary information due to the spatio-temporal heterogeneity of NDSI change, and constructs a spatio-temporal extreme gradient boosting model to recover the missing NDSI information. The method can reconstruct spatio-temporally continuous NDSI data, and the reconstruction result is accurate. Since only auxiliary data and spatio-temporal auxiliary data need to be input in the process of predicting NDSI, the method can still stably reconstruct data in the case of continuous NDSI loss. BRIEF DESCRIPTION OF DRAWINGS
[0031] The present application will be further described below with reference to the accompanying drawings and embodiments, wherein: The present application will be further described below with reference to the accompanying drawings and embodiments, wherein:
[0032] Figure 1 is a flow chart of a NDSI remote sensing data reconstruction method in an embodiment of the present application;
[0033] Figure 2 is a framework diagram of reconstructing NDSI remote sensing data based on a spatio-temporal gradient boosting model in an embodiment of the present application;
[0034] Figure 3 is a structural schematic diagram of a NDSI remote sensing data reconstruction device in an embodiment of the present application.
[0035] Figure 4 is a physical structure schematic diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION
[0036] In order to have a clearer understanding of the technical features, objectives and effects of the present application, the specific embodiments of the present application will be described in detail with reference to the drawings.
[0037] Please refer to Figure 1 and Figure 2 In an embodiment of the present application, a NDSI remote sensing data reconstruction method is provided, which specifically includes the following steps:
[0038] Step S1: Obtain NDSI data of the Greenland region, and obtain variable data related to the NDSI data as auxiliary data (Ancillary data), such as Figure 2 As shown in the figure, the auxiliary data includes elevation data (DEM), solar zenith angle data (SoA), solar azimuth angle data (SoZ), sensor zenith angle data (SeA), sensor azimuth angle data (SeZ), land cover type remote sensing data (LAC) and surface albedo remote sensing data (Albedo), and the spatial resolutions of the remote sensing data are unified.
[0039] Specifically, the remote sensing data of NDSI is derived from the existing remote sensing data set MOD10A1, and the spatial resolution is 500m. The remote sensing data of elevation is derived from the existing remote sensing data set Greenland DEM, and the spatial resolution is 500m. The remote sensing data of land cover type is derived from the existing remote sensing data set MCD12Q1, and the spatial resolution is 500m. The remote sensing data of sensor zenith angle, solar zenith angle, sensor azimuth angle and solar azimuth angle is derived from the existing remote sensing data set MOD09GA, and the spatial resolution is 1000m. The remote sensing data of surface albedo is derived from the existing remote sensing data set MOD10A1, and the spatial resolution is 500m. It should be particularly noted that, since the original surface albedo of MOD10A1 product is missing, the missing data is reconstructed by combining the spatio-temporal and physical models.
[0040] When unifying the spatial resolution of each remote sensing data, the remote sensing data of sensor zenith angle, solar zenith angle, sensor azimuth angle and solar azimuth angle are resampled to have the same spatial resolution as the remote sensing data of NDSI.
[0041] It should be understood that all remote sensing data can also be resampled as needed to unify them to a smaller spatial resolution.
[0042] Step S2: Extracting the spatio-temporal auxiliary data of each pixel, such as Figure 2 As shown, including temporal information and spatial information, specifically, the temporal information includes the time (day) corresponding to each pixel and the NDSI information of adjacent time effective pixels (TN), and the spatial information includes the longitude (Lat) and latitude (Lon) of each pixel and the NDSI information of adjacent spatial effective pixels (SN).
[0043] Step S3: Constructing a spatio-temporal extreme gradient boosting model, which can be formulated as:
[0044] NDSI = f(Day, Lat, Lon, DEM, LAC, SoA, SoZ, SeA, SeZ, Albedo, SN, TN)
[0045]
[0046]
[0047] Where Day represents the day of the year, Lat and Lon represent the longitude and latitude respectively, DEM represents the elevation, LAC represents the land cover type, SoA, SoZ, SeA, SeZ represent the zenith and azimuth of the sun, the zenith and azimuth of the sensor respectively, Albedo represents the surface albedo, (m, n) represents the spatial position of the pixel, t represents the sequence number of the pixel at time t, num represents the number of effective pixels, y m,n,t represents the NDSI value of the spatial position (m, n) on the tth day, SN represents the NDSI information of adjacent spatial effective pixels, TN represents the NDSI information of adjacent time effective pixels, and NDSI represents the predicted NDSI value.
[0048] Step S4: Inputting the NDSI data obtained in step S1, the auxiliary data and the spatio-temporal auxiliary data obtained in step S2 to train and test the model. As a preferred, the existing NDSI data and its corresponding auxiliary data and spatio-temporal auxiliary data are randomly divided into 70% for training the model and the remaining 30% for testing the model.
[0049] Step S5: inputting the auxiliary data corresponding to the missing pixel and the spatio-temporal auxiliary data into the model trained and tested in step S4 to recover the missing NDSI information.
[0050] The application fully utilizes the variables related to NDSI, further introduces spatio-temporal auxiliary information due to the spatio-temporal heterogeneity of NDSI change, and constructs a spatio-temporal extreme gradient boosting model to recover the missing NDSI information. The method can reconstruct the spatio-temporally continuous NDSI data, and the reconstruction result is high in accuracy. Since only auxiliary data and spatio-temporal auxiliary data need to be input in the process of predicting NDSI, the method can still robustly reconstruct data in the case of continuous missing of NDSI, such as Figure 2 The above method of the application finally reconstructs the seamless Greenland NDSI data.
[0051] In order to verify the performance of the spatio-temporal extreme gradient boosting (STXGBoost) model, the classical regression models (including multiple linear regression (MLR), K-Nearest Neighbor (KNN) and XGBoost model) are compared. In addition, by adding spatial auxiliary data and temporal auxiliary data in the XGBoost model, a spatial extreme gradient boosting (SXGBoost) model and a temporal extreme gradient boosting (TXGBoost) model are established respectively to verify the necessity of adding spatio-temporal information.
[0052] Taking the data of Greenland from the 155th day to the 165th day in 2020 as an example, 70% is randomly selected as training data, and the remaining 30% is selected as test data. Table 1 shows the prediction ability of different models.
[0053] Table 1. Quantitative evaluation of different models for predicting NDSI data
[0054]
[0055] The R 2 value of the MLR model is the lowest, and the RMSE is the highest, so obviously this model is not suitable for the reconstruction of NDSI data. The performance of the KNN model has improved, but it is still not accurate enough. In contrast, the XGBoost series of models shows excellent prediction ability. Among them, the SXGBoost model and the TXGBoost model respectively improve the accuracy of the predicted data by considering the spatial variability and the temporal variability. In general, the STXGBoost model performs best in R 2 , RMSE and MAE, which are 0.953, 0.031 and 0.011 respectively, indicating that combining spatio-temporal information can improve the prediction ability of the model and generate more accurate NDSI data.
[0056] The following describes an NDSI remote sensing data reconstruction device provided by the present invention. The NDSI remote sensing data reconstruction device described below can be referred to in correspondence with the NDSI data reconstruction method described above.
[0057] like Figure 3 As shown, an NDSI remote sensing data reconstruction device 100 in an embodiment of the present invention includes the following modules:
[0058] Data acquisition module 110 is used to acquire NDSI data of the area to be reconstructed, acquire variable data related to NDSI data as auxiliary data, and unify the spatial resolution of NDSI data and auxiliary data;
[0059] The data extraction module 120 is used to extract the spatiotemporal auxiliary data of each pixel;
[0060] Model building module 130 is used to build spatiotemporal extreme gradient boosting models;
[0061] The training and testing module 140 is used to train and test the model by inputting NDSI data, auxiliary data, and spatiotemporal auxiliary data.
[0062] The data reconstruction module 150 is used to input the auxiliary data and spatiotemporal auxiliary data corresponding to the missing pixels into the model after training and testing to recover the missing NDSI information.
[0063] like Figure 4 The diagram illustrates the physical structure of an electronic device, which may include a processor 610, a communication interface 620, a memory 630, and a communication bus 640. The processor 610, communication interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute the steps of the aforementioned NDSI remote sensing data reconstruction method. Specifically, this includes: acquiring elevation data, solar zenith angle data, solar azimuth angle data, sensor zenith angle data, sensor azimuth angle data, land cover type remote sensing data, and surface albedo remote sensing data as auxiliary data; extracting the time (number of days) corresponding to each pixel and the NDSI information of adjacent valid pixels as temporal auxiliary data; extracting the longitude, latitude, and NDSI information of adjacent valid pixels as spatial auxiliary data; and constructing a spatiotemporal extreme gradient boosting model, which, when inputting the auxiliary data corresponding to the missing pixels and the spatiotemporal auxiliary data, can recover the missing NDSI information.
[0064] Further, the logic instructions in the memory 630 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.
[0065] In another aspect, the embodiments of the present application also provide a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the NDSI remote sensing data reconstruction method described above, and specifically includes: obtaining elevation data, solar zenith angle data, solar azimuth angle data, sensor zenith angle data, sensor azimuth angle data, land cover type remote sensing data, and surface albedo remote sensing data as auxiliary data; extracting time (days) corresponding to each pixel and NDSI information of adjacent time effective pixels as time auxiliary data; extracting longitude and latitude of each pixel and NDSI information of adjacent spatial effective pixels as spatial auxiliary data; constructing a spatio-temporal extreme gradient boosting model, and inputting auxiliary data and spatio-temporal auxiliary data corresponding to the missing pixel to recover the missing NDSI information.
[0066] The NDSI remote sensing data reconstruction method, device, equipment and storage medium provided by the embodiments of the present application make full use of the variable information related to NDSI, and consider the spatio-temporal variability, so that the spatio-temporally continuous NDSI data is reconstructed efficiently and accurately, and support is provided for climate change research and water resource management.
[0067] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusion, so that processes, methods, articles or systems that include a series of elements not only include those elements, but also include other elements not explicitly listed, or inherent to such processes, methods, articles or systems. Without more limitations, the element defined by the statement "including a" does not exclude the presence of additional identical elements in the process, method, article or system including the element.
[0068] The above-mentioned embodiment serial numbers of the application are only for description, and do not represent the advantages and disadvantages of the embodiments. In the unit claims in which several devices are listed, several of the devices can be embodied by the same hardware item. The use of the words first, second, and third does not represent any order, and the words can be interpreted as identifiers.
[0069] The above is only the preferred embodiment of the application, and does not limit the patent scope of the application. Any equivalent structure or equivalent process transformation using the content of the application specification and drawings, or direct or indirect application in other related technical fields, is also included in the patent protection scope of the application.
Claims
1. A method for reconstructing NDSI remote sensing data, characterized in that, The method comprises the following steps: acquiring NDSI data of a region to be reconstructed, acquiring variable data related to the NDSI data as auxiliary data, and unifying the spatial resolutions of the NDSI data and the auxiliary data; extracting time-space auxiliary data of each pixel; constructing a time-space extreme gradient boosting model; inputting the NDSI data, the auxiliary data, and the time-space auxiliary data to train and test the model; inputting auxiliary data and time-space auxiliary data corresponding to missing pixels into the trained and tested model to recover the missing NDSI information; the auxiliary data comprises elevation data, solar zenith angle data, solar azimuth angle data, sensor zenith angle data, sensor azimuth angle data, land cover type remote sensing data, and surface albedo remote sensing data; the time-space auxiliary data comprises time auxiliary data and space auxiliary data; the time auxiliary data comprises time corresponding to each pixel and NDSI information of adjacent time effective pixels; and the space auxiliary data comprises longitude and latitude of each pixel and NDSI information of adjacent space effective pixels; the time-space extreme gradient boosting model can be formulated as: wherein, denotes the day of the year, and denote the longitude and latitude, respectively, denotes the elevation, denotes the land cover type, denote the solar zenith and azimuth, respectively, and the sensor's zenith and azimuth, denotes the surface albedo, denotes the spatial position of a pixel, t denotes the time series number of the pixel at time t, and num denotes the number of valid pixels, denotes the NDSI value of the pixel at spatial position at day t, SN denotes the NDSI information of the neighboring spatial valid pixels, TN denotes the NDSI information of the neighboring temporal valid pixels, and NDSI denotes the predicted NDSI value.
2. The NDSI remote sensing data reconstruction method of claim 1, wherein, the time-space extreme gradient boosting model inputs data including the auxiliary data and the time-space auxiliary data, and 70% of existing NDSI data and corresponding auxiliary data and time-space auxiliary data are randomly divided for training the model, and the remaining 30% is used for testing the model.
3. An NDSI remote sensing data reconstruction apparatus, characterized in that, The method comprises the following modules: a data acquisition module, configured to acquire NDSI data of a region to be reconstructed, acquire variable data related to the NDSI data as auxiliary data, and unify the spatial resolutions of the NDSI data and the auxiliary data; a data extraction module, configured to extract time-space auxiliary data of each pixel; a model construction module, configured to construct a time-space extreme gradient boosting model; a training and testing module, configured to input the NDSI data, the auxiliary data, and the time-space auxiliary data to train and test the model; a data reconstruction module, configured to input auxiliary data and time-space auxiliary data corresponding to missing pixels into the trained and tested model to recover the missing NDSI information; the auxiliary data comprises elevation data, solar zenith angle data, solar azimuth angle data, sensor zenith angle data, sensor azimuth angle data, land cover type remote sensing data, and surface albedo remote sensing data; the time-space auxiliary data comprises time auxiliary data and space auxiliary data; the time auxiliary data comprises time corresponding to each pixel and NDSI information of adjacent time effective pixels; and the space auxiliary data comprises longitude and latitude of each pixel and NDSI information of adjacent space effective pixels; the time-space extreme gradient boosting model can be formulated as: wherein, denotes the day of the year, and denote the longitude and latitude, respectively, denotes the elevation, denotes the land cover type, denote the solar zenith and azimuth, respectively, the sensor's zenith and azimuth, denotes the surface albedo, denotes the spatial position of a pixel, t denotes the time series number of the pixel at time t, and num denotes the number of valid pixels, denotes the NDSI value of the pixel at spatial position on day t, SN denotes the NDSI information of the neighboring spatial valid pixels, TN denotes the NDSI information of the neighboring temporal valid pixels, and NDSI denotes the predicted NDSI value.
4. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, the processor implements the steps of the NDSI remote sensing data reconstruction method according to any one of claims 1-2 when executing the program.
5. A storage medium having stored thereon a computer program, characterized in that the computer program implements the steps of the NDSI remote sensing data reconstruction method according to any one of claims 1-2 when executed by the processor.