Method and device for simulating reflectivity of ground objects based on inter-band intrinsic spectral characteristics

Abstract

The application discloses a ground object reflectivity simulation method and device based on inter-band inherent spectral characteristics, and belongs to the technical field of remote sensing science. The method firstly acquires hyperspectral data and MODIS reflectivity product MOD09A1 data, and performs spatial matching with GMTED2010 data; secondly, the data are preprocessed, including projection conversion, resampling, multi-layer integration, MODIS multispectral data simulation, spectral index calculation and sample production, so as to construct a training data set and a prediction data set; then, hyperparameter optimization is performed based on a random forest model, a regression model between the reflectivity of seven visible-near infrared bands of the MOD09A1 data and the reflectivity of a target band (such as 532 nm) is established; and finally, the trained model is used to simulate the reflectivity of unobserved bands of a satellite. The application innovatively uses inter-band inherent spectral characteristics, solves the problem of missing of the reflectivity data of unobserved bands of the satellite, and has high simulation precision.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of remote sensing science and technology, and particularly relates to a ground object reflectivity simulation method and device based on inter-band intrinsic spectral characteristics. BACKGROUND

[0002] Ground surface reflectivity is a key parameter for characterizing the energy balance, material cycle and ecological environment state of the earth's surface, and plays an important role in environmental monitoring, disaster warning and climate change research. Active remote sensing technologies such as lidar rely on ground surface reflectivity data at specific bands (such as 532 nm and 1064 nm), but existing satellite sensors cannot directly obtain observation data at these bands, which seriously restricts the development of related applications.

[0003] Currently, ground surface reflectivity retrieval methods mainly include atmospheric correction based on physical models and band relationship modeling based on statistical learning. For example, satellite data such as MODIS can retrieve ground surface reflectivity at observed bands through atmospheric correction algorithms, but for unobserved bands (such as lidar operating bands), existing methods have obvious deficiencies: (1) Although hyperspectral data (such as AVIRIS) can provide continuous spectral information, the coverage is limited and the update frequency is low, making it difficult to meet the needs of large-scale dynamic monitoring; (2) Traditional statistical models are mostly designed for specific bands, and lack of systematic exploration of inter-band intrinsic spectral characteristics, resulting in insufficient simulation accuracy of unobserved band reflectivity.

[0004] In recent years, machine learning technology has provided a new approach to band relationship modeling, but existing research has mostly focused on reflectivity retrieval of observed bands, and there is still a lack of effective methods for simulating unobserved bands. Therefore, there is an urgent need for a high-precision reflectivity simulation method that can fully utilize the advantages of multi-source data and explore the spectral correlation between bands, to fill the gap in satellite unobserved band reflectivity data. SUMMARY

[0005] To solve the above technical problems, the application provides a ground object reflectivity simulation method and device based on inter-band intrinsic spectral characteristics, which uses the spectral correlation of adjacent bands of typical ground objects to establish a regression relationship between the reflectivity of the 7 visible-near infrared bands of MOD09A1 data and the reflectivity of unobserved bands of satellites, and realizes the simulation of the ground surface reflectivity of the band.

[0006] To achieve the above purpose, the technical solution adopted by the application is as follows:

[0007] A ground object reflectivity simulation method based on inter-band intrinsic spectral characteristics, the method comprising:

[0008] Step 1, taking hyperspectral data and MODIS reflectance product MOD09A1 data as the benchmark, GMTED2010 data corresponding to the same time and space is obtained respectively;

[0009] Step 2, pre-processing operation is performed on the hyperspectral data and GMTED2010 data to make a training data set;

[0010] Step 3, pre-processing operation is performed on the MODIS reflectance product MOD09A1 data and GMTED2010 data to make a prediction data set;

[0011] Step 4, the training data set is input into a random forest model for training, hyperparameter optimization is performed, and a surface feature reflectance simulation model is obtained, the surface feature reflectance simulation model being a regression model between the reflectance of 7 visible-near infrared bands of the MOD09A1 data and the reflectance of a target unobserved band;

[0012] Step 5: the prediction data set is input into the surface feature reflectance simulation model, and the reflectance data of the target band is output.

[0013] On the other hand, the application provides a surface feature reflectance simulation device based on the inherent spectral characteristics between bands, comprising:

[0014] A data acquisition module is configured to take hyperspectral data and MODIS reflectance product MOD09A1 data as the benchmark, and GMTED2010 data corresponding to the same time and space is obtained respectively;

[0015] A training data set module is configured to perform pre-processing operation on the hyperspectral data and GMTED2010 data to make a training data set;

[0016] A prediction data set module is configured to perform pre-processing operation on the MODIS reflectance product MOD09A1 data and GMTED2010 data to make a prediction data set;

[0017] A simulation model module is configured to input the training data set into a random forest model for training, perform hyperparameter optimization, and obtain a surface feature reflectance simulation model, the surface feature reflectance simulation model being a regression model between the reflectance of 7 visible-near infrared bands of the MOD09A1 data and the reflectance of a target unobserved band;

[0018] A reflectance calculation module is configured to input the prediction data set into the surface feature reflectance simulation model, and output the reflectance data of the target band.

[0019] In a third aspect, the present application provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned land surface reflectance simulation method based on inter-band intrinsic spectral features.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, enable the processor to implement the aforementioned land surface reflectance simulation method based on inter-band intrinsic spectral features.

[0021] The present application has the following beneficial effects:

[0022] The present application fills the gap of satellite unobserved band reflectance data, and for the first time realizes effective simulation of satellite unobserved band reflectance by establishing a high-precision regression model between the 7 visible-near infrared bands of MOD09A1 data and the target band (such as 532nm), solving the key data missing problem in the application field of laser radar and other fields.

[0023] The simulation accuracy is high, and the reliability is strong. Based on the random forest model and the inter-band spectral feature optimization, the determination coefficient R² of the simulation result and the true observation value reaches 0.9686, and the root mean square error (RMSE) is only 0.0200, which is significantly better than the traditional statistical method, and meets the high-precision application demand.

[0024] The spatio-temporal coverage capability is outstanding. Relying on the advantages of global coverage and 8-day time resolution of MODIS data, a high-frequency updated reflectance data set in the global range can be produced, breaking through the bottleneck of limited spatio-temporal coverage of airborne data.

[0025] The degree of automation is high, and the applicability is wide. The standardized pretreatment process and automatic model training are adopted, which can quickly adapt to different regional and band requirements, and provide flexible data support for surface monitoring, disaster warning and other applications.

[0026] Multi-source data is cooperatively utilized. The spectral detail advantage of hyperspectral data and the spatio-temporal coverage advantage of MODIS data are innovatively integrated, and the accuracy of reflectance simulation in complex environment is significantly improved through terrain data assisted correction. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 A schematic diagram of the land surface reflectance simulation method based on inter-band intrinsic spectral features of the present application;

[0028] Figure 2 A flowchart of the land surface reflectance simulation method based on inter-band intrinsic spectral features of the present application;

[0029] Figure 3This is a schematic diagram of the random forest model structure;

[0030] Figure 4 Accuracy map of 532nm band reflectance for ground object reflectance model simulation. Detailed Implementation

[0031] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0032] like Figure 1 and Figure 2 The diagram shown is a flowchart of the ground cover reflectance simulation method based on inherent spectral characteristics between bands according to the present invention, including:

[0033] Step 1: Using hyperspectral data and MODIS reflectance product MOD09A1 data as a reference, obtain the corresponding GMTED2010 data in time and space.

[0034] The hyperspectral data used in this application was acquired by the AVIRIS airborne hyperspectral sensor and is currently freely available to the public. The download website is https: / / avirisng.jpl.nasa.gov / dataportal / . The data covers a spectrum from 376 to 2500 nm with a spectral resolution of 5 nm, comprising 425 bands. The reflectance corresponding to seven visible-near-infrared bands of the MOD09A1 data was obtained as input for model training, and the reflectance at 532 nm was used as the ground truth for model training. The MOD09A1 data used in this application is an 8-day composite product of surface reflectance collected by the MODIS sensor aboard the Terra satellite, containing seven visible-near-infrared bands (Band 1: 620–670 nm, Band 2: 841–876 nm, Band 3: 459–479 nm, Band 4: 545–565 nm, Band 5: 1230–1250 nm, Band 6: 1628–1652 nm, Band 7: 2105–2155 nm). The reflectance of these seven visible-near-infrared bands was obtained as input for model prediction.

[0035] The GMTED2010 data used in this application is global topographic elevation data from 2010. Complex terrain alters the radiative transfer process of solar radiation energy between the sun, the earth's surface, and the sensor, resulting in different reflectance values ​​for the same ground feature in the same image at different slopes and aspects. Using elevation values ​​as input parameters for ground feature reflectance simulation models allows for more accurate simulation of reflectance. Therefore, this application subsequently performs spatial matching of the GMTED2010 data with hyperspectral data and MOD09A1 data, selecting GMTED2010 data with the same latitude and longitude values ​​as the data to be matched, and using them as input for model training / prediction.

[0036] Step 2: Preprocessing operation for hyperspectral data and GMTED2010 data, making training data set.

[0037] The preprocessing operation of the present application includes hyperspectral data projection conversion, GMTED2010 data resampling, hyperspectral data and GMTED2010 data multi-layer integration, MODIS multispectral data simulation, and spectral index calculation.

[0038] The hyperspectral data projection conversion includes: the hyperspectral data is projected in UTM_Zone_11N, the projection of the hyperspectral data is converted to GCS_WGS_1984 by software, the resampling technique is set to "BILINEAR", and the output pixel size is set to 0.00416 degrees (about 500 meters).

[0039] The GMTED2010 data resampling includes: the spatial resolution of the GMTED2010 data is 250 m, and the spatial resolution of the GMTED2010 data is resampled to 0.00416 degrees (about 500 meters) by software.

[0040] The hyperspectral data and GMTED2010 data multi-layer integration includes: the bands of the hyperspectral data and the GMTED2010 data are combined into a new raster data set, and each integrated image contains 46 bands.

[0041] The MODIS multispectral data simulation includes: MOD09A1 is multispectral data, and the reflectance data of 7 wide bands are recorded by the moderate resolution imaging spectrometer MODIS, while the hyperspectral data record the reflectance at multiple adjacent (5 nm interval) wavelengths. Therefore, the hyperspectral reflectance needs to be converted to the equivalent reflectance of the 7 visible-near infrared bands of the MOD09A1 data by the spectral response function.

[0042] The formula for converting the hyperspectral reflectance to the multispectral reflectance is as follows:

[0043] ,

[0044] In the formula, is the reflectance of the calculated th wide band (7 visible-near infrared bands of the MOD09A1 data), is the reflectance at wavelength , is the spectral response of the th wide band of the MOD09A1 data at wavelength , is the lower limit value of the wide band spectral range, The upper limit value of the wide-band spectral range.

[0045] The spectral index calculation includes calculating the following four spectral features on the equivalent reflectance, which helps the model learn the change rule between the reflectance of the seven visible-near infrared bands of MOD09A1 data and the reflectance of the satellite unobserved bands in different surface feature types.

[0046] (a) The normalized vegetation index (NDVI) can effectively reflect the growth condition of vegetation and the vegetation coverage, and is beneficial to distinguish vegetation and water body, construction land, bare land, snow and other non-vegetation information. Its calculation formula is:

[0047] ,

[0048] In the formula, is the near-infrared band reflectance, is the infrared band reflectance, which is Band2 and Band1 of the simulated equivalent reflectance respectively, and the center wavelengths are 858nm and 645nm respectively.

[0049] (b) The normalized water index (NDWI) can be used to quickly, simply and accurately extract water information. Its calculation formula is:

[0050] ,

[0051] In the formula, is the green band reflectance, is the near-infrared band reflectance, which is Band4 and Band2 of the simulated equivalent reflectance respectively, and the center wavelengths are 555nm and 858nm respectively.

[0052] (c) The bare soil index (BSI) can enhance the brightness value of bare land, and effectively distinguish bare land from vegetation, water body, construction land, snow and other non-bare land information. Its calculation formula is:

[0053] ,

[0054] In the formula, , , , are the mid-infrared band, infrared band, near-infrared band and blue band reflectance respectively, corresponding to Band6, Band1, Band2 and Band3 of the simulated equivalent reflectance, and the center wavelengths are 1640nm, 645nm, 858nm and 469nm respectively.

[0055] (d) Normalized Difference Barren Index (NDBI) can accurately reflect the information of building land, and the greater the value, the higher the proportion of building land and the higher the building density. Its calculation formula is:

[0056] ,

[0057] In the formula, is the mid-infrared band reflectivity, is the near-infrared band reflectivity, which is Band6 and Band2 of the simulated equivalent reflectivity respectively, and the center wavelengths are 1640nm and 858nm respectively.

[0058] Step 3: Preprocessing operation is performed on MOD09A1 data and GMTED2010 data to make prediction data set.

[0059] The preprocessing operation of the present application includes MOD09A1 data preprocessing, GMTED2010 data resampling, multi-layer integration of MOD09A1 data and GMTED2010 data, spectral index calculation, and sample making.

[0060] MOD09A1 data preprocessing includes: the MOD09A1 data downloaded from the United States Aerospace Administration is in HDF format, and the original projection is "sine curve projection". The MOD09A1 data is converted from HDF format to GeoTIFF format by ENVI software, and the projection is converted to GCS_WGS_1984.

[0061] GMTED2010 data resampling includes: the spatial resolution of GMTED2010 data is 250m, and the spatial resolution of GMTED2010 data is resampled to 0.00416 degrees (about 500 meters) by software.

[0062] Multi-layer integration of MOD09A1 data and GMTED2010 data includes: merging the bands of MOD09A1 data and GMTED2010 data into a new raster data set, and each integrated image contains 8 bands.

[0063] Spectral index calculation includes: calculating normalized vegetation index NDVI, normalized water index NDWI, bare soil index BSI, and normalized building index NDBI on the reflectivity of MOD09A1 data.

[0064] Sample making includes: the multi-layer integrated data contains various information, and some conditions need to be set to filter useful information, clean the data, and ensure the quality of the data. The specific steps are as follows:

[0065] (a) Determining the effective value range;

[0066] MODIS itself is not a dedicated remote sensor for water color. Under the seawater type, the surface reflectance can be negative in all bands. Therefore, when making samples, it is necessary to filter each band of the MOD09A1 data and only keep the samples whose values ​​are within the normal range of surface reflectance (between 0 and 1).

[0067] (b) Determination of cloud cover;

[0068] Due to abundant water vapor supply, plentiful rainfall, and significant vertical climate in some areas, cloud cover exists year-round. However, cloud cover hinders satellite observation of the land surface, preventing the direct provision of accurate land surface reflectance. Therefore, when creating prediction samples, it is necessary to remove cloud-covered pixels from the MOD09A1 data based on its quality control bands.

[0069] Step 4: Construct a random forest model. Use the training dataset as input features to train the random forest model, perform hyperparameter optimization, and obtain a ground reflectance simulation model. The ground reflectance simulation model is a regression model between the reflectance of the seven visible-near infrared bands of MOD09A1 data and the reflectance of the unobserved bands of the target.

[0070] like Figure 3 As shown, the regression model here refers to the linear relationship between the reflectance of the seven visible-near-infrared bands of the MOD09A1 data and the reflectance of the unobserved bands of the target. In the model training phase, the hyperspectral data, after spectral convolution preprocessing, is equivalent to the reflectance of the seven visible-near-infrared bands of the MOD09A1 data, which is used as the input for model training. Then, the reflectance of the hyperspectral data at 532nm is obtained as the ground truth for model training. This model training process can be viewed as learning the linear relationship between the reflectance of the seven visible-near-infrared bands of the MOD09A1 data and the 532nm reflectance. In the prediction phase, the reflectance of the seven visible-near-infrared bands of the MOD09A1 data is input, and the 532nm reflectance can be calculated based on this linear relationship.

[0071] During model building, some parameters need to be set in advance. Hyperparameters are parameters that are pre-defined before model training begins, and their proper configuration plays a crucial role in the model's performance. Grid search is one of the most widely used hyperparameter optimization methods. This method predefines a list of candidate hyperparameter values ​​and performs a comprehensive exhaustive search of all possible parameter combinations. Then, cross-validation is used to evaluate the performance of each parameter combination, and finally, the optimal parameter combination is selected as the model's hyperparameters.

[0072] The application constructs a random forest model, uses a grid search method to gradually optimize the hyperparameters of the model by taking the training data set as the input of the random forest model, and finds the optimal model performance hyperparameter combination through five-fold cross-validation. The random forest model is retrained using the optimal hyperparameter combination to establish the mapping relationship between the wave bands; the model performance is evaluated on the validation set to ensure that the precision requirement is met. The hyperparameters of the random forest model are the number of decision trees n_estimators, the maximum depth of the decision tree max_depth, the minimum number of samples that can be split min_samples_split, and the minimum number of samples required for the leaf node min_samples_leaf. The search range of the hyperparameters of the random forest model and the optimal hyperparameters are shown in Table 1.

[0073] Table 1

[0074]

[0075] Step 5: Input the prediction data set into the ground feature reflectivity simulation model to simulate the reflectivity of the satellite unobserved wave band, and output the reflectivity data of the target wave band.

[0076] Select a certain area, simulate the 532nm wave band reflectivity based on the trained ground feature reflectivity simulation model.

[0077] When simulating the 532nm wave band reflectivity, the reflectivity of the MODIS fourth wave band (wavelength range: 545-565nm) closest to the 532nm wave band is taken as the true value. Finally, the model-simulated reflectivity and the MODIS satellite-observed reflectivity are calculated to determine the coefficient R 2 , root mean square error RMSE, and mean absolute error MAE, as shown in Figure 4 . The horizontal axis of each graph represents the MODIS satellite-observed wave band reflectivity, and the vertical axis represents the model-simulated reflectivity. The scatter points in the graph represent the data points. In the scale axis on the right side of the graph, the black color at the bottom end represents a lower scatter point density, and the gray-white color in the middle represents a higher scatter point density. A 1:1 standard line (dotted line) and a fitting line (solid line) are also plotted in the graph to evaluate the accuracy of the model simulation. Most of the data points in the graph are basically distributed near the 1:1 standard line (dotted line), indicating that the model has high prediction accuracy for the 532nm reflectivity.

[0078] In summary, the application effectively solves the problem of missing satellite unobserved wave band reflectivity data by using an innovative multi-source data fusion strategy and machine learning modeling method, and provides reliable technical support for remote sensing quantitative inversion and laser radar application. The experimental data fully verify the high precision, high stability and good engineering applicability of the method.

[0079] In another aspect, the present application provides a ground object reflectivity simulation device based on inter-band inherent spectral characteristics, which comprises various modules capable of realizing various steps of the aforementioned method, and specifically comprises:

[0080] A data acquisition module is configured to acquire GMTED2010 data corresponding to hyperspectral data and MODIS reflectivity product MOD09A1 data in time and space respectively based on the hyperspectral data and the MODIS reflectivity product MOD09A1 data.

[0081] A training data set module is configured to perform preprocessing operations on the hyperspectral data and the GMTED2010 data to make a training data set.

[0082] A prediction data set module is configured to perform preprocessing operations on the MODIS reflectivity product MOD09A1 data and the GMTED2010 data to make a prediction data set.

[0083] A simulation model module is configured to input the training data set into a random forest model for training, perform hyperparameter optimization, and obtain a ground object reflectivity simulation model, which is a regression model between reflectivity of 7 visible-near infrared bands of the MOD09A1 data and reflectivity of a target unobserved band.

[0084] A reflectivity calculation module is configured to input the prediction data set into the ground object reflectivity simulation model and output reflectivity data of a target band.

[0085] In a third aspect, the present application provides an electronic device, comprising: one or more processors; a memory configured to store one or more programs; and wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the aforementioned ground object reflectivity simulation method based on inter-band inherent spectral characteristics.

[0086] In a fourth aspect, the present application provides a computer-readable storage medium having stored executable instructions, which, when executed by a processor, enable the processor to implement the aforementioned ground object reflectivity simulation method based on inter-band inherent spectral characteristics.

[0087] The above-described specific embodiments further illustrate the purposes, technical solutions and beneficial effects of the present application, and it should be understood that the above-described specific embodiments are merely specific embodiments of the present application and are not intended to limit the present application, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for simulating the reflectance of ground objects based on inherent spectral characteristics between bands, characterized in that, The method includes: Step 1: Using hyperspectral data and MODIS reflectance product MOD09A1 data as a reference, obtain the corresponding GMTED2010 data in time and space. Step 2: Preprocess the hyperspectral data and GMTED2010 data to create a training dataset; Step 3: Perform preprocessing operations on MODIS reflectivity product MOD09A1 data and GMTED2010 data to create a prediction dataset; Step 4: Input the training dataset into the random forest model for training, perform hyperparameter optimization, and obtain the ground reflectance simulation model. The ground reflectance simulation model is a regression model between the reflectance of the 7 visible-near infrared bands of MOD09A1 data and the reflectance of the unobserved bands of the target. Step 5: Input the predicted dataset into the ground reflectance simulation model and output the reflectance data of the target band.

2. The method for simulating ground reflectance based on inherent spectral characteristics between bands according to claim 1, characterized in that, In step 1, the hyperspectral data has a spectral coverage range of 376–2500 nm and a spectral resolution of 5 nm; the MOD09A1 data contains 7 visible-near-infrared bands; and the GMTED2010 data is topographic elevation data.

3. The method for simulating ground reflectance based on inherent spectral characteristics between bands according to claim 1, characterized in that, In step 2, the preprocessing operations include hyperspectral data projection conversion, GMTED2010 data resampling, multi-layer integration of hyperspectral data and GMTED2010 data, MODIS multispectral data simulation, and spectral index calculation.

4. The method for simulating ground reflectance based on inherent spectral characteristics between bands according to claim 3, characterized in that, The spectral index calculation includes the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Bare Soil Index (BSI), and Normalized Building Index (NDBI).

5. The method for simulating ground reflectance based on inherent spectral characteristics between bands according to claim 1, characterized in that, In step 3, the preprocessing operations include MOD09A1 data preprocessing, GMTED2010 data resampling, multi-layer integration of MOD09A1 and GMTED2010 data, spectral index calculation, and sample preparation.

6. The method for simulating ground reflectance based on inherent spectral characteristics between bands according to claim 1, characterized in that, Step 4 includes: Initialize the random forest regression model and define the set of hyperparameters to be optimized; Configure the parameter space range for grid search, including the number of decision trees and the maximum depth; A grid search optimization strategy using five-fold cross-validation is employed to determine the optimal hyperparameter combination. The random forest model is retrained using the optimal hyperparameter combination to establish the mapping relationship between bands; Evaluate the model performance on the validation set to ensure that the accuracy requirements are met.

7. The method for simulating ground reflectance based on inherent spectral characteristics between bands according to claim 5, characterized in that, The sample preparation includes screening each band of the MOD09A1 data, retaining only samples whose surface reflectance values ​​are between 0 and 1; and removing cloud-covered pixels from the MOD09A1 data based on the quality control bands of the MOD09A1 data.

8. A ground cover reflectance simulation device based on inherent spectral characteristics between bands, characterized in that, include: The data acquisition module is used to acquire GMTED2010 data corresponding to hyperspectral data and MODIS reflectance product MOD09A1 data in time and space, respectively, based on hyperspectral data and MODIS reflectance product MOD09A1 data. The training dataset module is used to preprocess hyperspectral data and GMTED2010 data to create the training dataset; The Predictive Dataset module is used to preprocess MODIS reflectivity product MOD09A1 data and GMTED2010 data to create a predictive dataset. The simulation model module is used to input the training dataset into the random forest model for training, perform hyperparameter optimization, and obtain a ground reflectance simulation model. The ground reflectance simulation model is a regression model between the reflectance of the seven visible-near infrared bands of the MOD09A1 data and the reflectance of the unobserved bands of the target. The reflectance calculation module is used to input the predicted dataset into the ground reflectance simulation model and output the reflectance data of the target band.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by the one or more processors, the one or more processors implement the ground reflectance simulation method based on the inherent spectral characteristics between bands as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, enable the processor to implement the ground reflectance simulation method based on the inherent spectral characteristics between bands as described in any one of claims 1-7.

Images