A method for normalizing reflectivity of multi-source remote sensing satellites

Abstract

The invention relates to the field of agricultural remote sensing technology, specifically a method for standardizing the reflectivity of a multi-source remote sensing satellite. The invention uses MODIS data as a reference data set, conducts reflectivity standardization research on satellites, uses PSF to simulate the characteristics and spectral response function of a MODIS sensor, models the PSF as a Gaussian distribution function, and obtains the best standard deviation; degenerates the reflectivity image to the spatial resolution of MODIS and establishes a correlation, applies a Bayesian framework to establish a statistical relationship between the standardized image and the original image reflectivity image, and between the standardized image and the MODIS image, thereby utilizing maximum a posteriori estimation. The invention significantly enhances the density on the time series image, smoothes and increases the details of the vegetation index time series curve, can make the monitoring of crops more refined, and can also provide support for precision agriculture, thereby optimizing the use of agricultural resources and improving agricultural production efficiency and sustainability.

Description

Technical Field

[0001] The invention relates to the technical field of agricultural remote sensing, in particular to a method for standardizing reflectivity of multi-source remote sensing satellites. Background Art

[0002] The healthy development of agriculture is directly related to the country's economic stability and social prosperity. Therefore, it is necessary to monitor key information such as crop growth, soil moisture, pests and diseases in a timely and accurate manner to provide decision support and refined management.

[0003] Remote sensing technology has the characteristics of rapid and non-destructive acquisition of ground information, and it is of great significance to use remote sensing technology to support agricultural development. Among them, high spatial resolution satellite data has rich spatial information and can better monitor various information of farmland. Although existing high spatial resolution data are used in crop monitoring research, they are limited by sensors, resulting in long revisit cycles and low frequency of repeated sampling observations in the same area. In addition, due to weather factors such as clouds and rain, the available remote sensing images may be further reduced. Therefore, the coordinated use of remote sensing satellites with multiple data sources is an important means to solve the above problems.

[0004] The coordinated use of multi-source remote sensing satellites can significantly improve the spatial and temporal resolution of data, achieving higher spatial accuracy and more frequent monitoring, which is crucial for phenological monitoring that changes rapidly and is densely detailed. In addition, the use of multiple data sources can also improve the accuracy and reliability of ground observations, reduce noise, cloud cover and other data quality issues by improving the independence and diversity of observation data, thereby enhancing the accuracy and reliability of data, thereby accelerating the acquisition of key information on crops and improving the accuracy of decision-making, which is of great significance to crop monitoring.

[0005] However, with the growth of more and more remote sensing data and the gradual improvement of international and domestic differences in spatial resolution, although this differentiation has a positive effect on the diversity of monitoring data, when different satellite data sources are used in coordination, different satellite data sensors have different spectral responses, band configurations and detection characteristics, which will lead to differences in reflectivity of data obtained by different sensors and make it impossible to form a unified standard, resulting in inconsistency in data in coordinated use, which in turn affects the precision and accuracy of the data.

[0006] Therefore, it is necessary to design a method for standardizing the reflectivity of multi-source remote sensing satellites to significantly enhance the density of time series images, smooth and increase the details of the vegetation index time series curve, make crop monitoring more refined, and provide support for precision agriculture, optimize the use of agricultural resources, and improve agricultural production efficiency and sustainability. Summary of the invention

[0007] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for standardizing the reflectivity of multi-source remote sensing satellites, which significantly enhances the density on time series images, smoothes and increases the details of the vegetation index time series curve, makes the monitoring of crops more refined, and further provides support for precision agriculture, optimizes the use of agricultural resources, and improves agricultural production efficiency and sustainability.

[0008] In order to achieve the above object, the present invention provides a method for standardizing the reflectivity of multi-source remote sensing satellites, comprising the following steps:

[0009] S1, obtain GF-1, GF-6 and Sentinel-2 satellite data, as well as MODIS data, and pre-process GF-1 and GF-6 data;

[0010] S2, taking MODIS data as the reference data set, uses PSF to simulate the imaging effect and spectral response function of MODIS, and degrades the data of GF-1, GF-6 and Sentinel-2 to the spatial resolution of MODIS to solve the problem of large difference in spatial resolution;

[0011] S3, based on the Bayesian framework, the statistical relationship between the standardized image and the original image reflectance image, as well as between the standardized image and the MODIS image, is established, so as to obtain the reflectance standardized image using the maximum a posteriori estimation, i.e., data with a spatial resolution of 15 m and the spectral response function characteristics of MODIS;

[0012] S4, a combination of qualitative and quantitative analysis methods was used to verify the accuracy of the reflectance of multi-source remote sensing satellite images based on a single remote sensing image before and after normalization and MODIS reflectance;

[0013] Among them, the qualitative analysis selects the true color difference of the visual inspection image; the quantitative analysis introduces the determination coefficient R2, the root mean square error RMSE and the relative root mean square error rRMSE for evaluation and analysis, among which the rRMSE is introduced to compare the RMSE between different data sets to obtain a more accurate evaluation;

[0014] Subsequently, a time series analysis was conducted on the satellite images before and after reflectivity standardization. At the same time, the verified reflectivity standardization product HLS data was introduced to conduct a time series comparative analysis with the images before and after standardization to prove the feasibility of this method.

[0015] The specific process of preprocessing in S1 is:

[0016] Download GF-1 and GF-6 images of the experimental area from the website of China Resources Satellite Application Center, download L2A-level products of Sentinel-2 data of the experimental area from the official website of ESA, and download MODIS MCD43A4 data and Harmonized Landsat Sentinel-2 (HLS) data from the official website of NASA; perform image preprocessing on GF-1 and GF-6 images, including radiation calibration, atmospheric correction, orthorectification and geometric registration, and use Sentinel-2 images as the reference image for geometric registration.

[0017] The specific steps of S2 are:

[0018] The point spread function is modeled as a Gaussian distribution curve to characterize the spectral response function. The reflectance data of GF-1, GF-6 and Sentinel-2 should be as similar as possible to MODIS when downgraded to MODIS resolution by considering the point spread function of the MODIS sensor. The formula is:

[0019] A 463m (λ)=subsamplepsf(λ)*A 15m (λ);

[0020] Among them: A 463m (λ) is the degraded GF-1, GF-6 and Sentinel-2 images at the MODIS 463.3127m resolution with a wavelength of λ; A 15m (λ) is the preprocessed 15.4438m resolution reflectance image of GF-1, GF-6 and Sentinel-2 with wavelength λ; PSF(λ) is the n×n convolution kernel matrix representing MODIS PSF, * represents convolution, subsample is the subsampling of every 30 pixels of the convolution image, where 30≈463.3127 / 15.4438;

[0021] The effective spatial resolution of MODIS images is about 600m to 1000m. Set n to 90, that is, use 90×90 pixels of high-resolution images for convolution, and model the PSF as a Gaussian function:

[0022]

[0023] where psf λ (i,j) are the filter weights with i and j∈(1,2,…,90), so i=46,j=46 is the center of the 90×90 filter;

[0024] σλ(m) is the standard deviation specific to the study area and band;

[0025] σλ It is empirically derived by maximizing the correlation coefficient between degraded GF-1, GF-6, Sentinel-2, and MODIS images, and is given by:

[0026]

[0027] R 2 For two variables A 463m (λ,σ λ ) and M 463m The coefficient of determination of (λ) is 463m (λ,σ λ ) as σ λ The function of M is derived through the point spread function. 463m (λ) represents the MODIS NBAR image with a resolution of 463.3127 m.

[0028] The specific steps of the Bayesian method in S3 are:

[0029] The point spread function is introduced to represent the prior distribution. The reflectance of the image after reflectance standardization is obtained by combining the spatial information of GF-1, GF-6 and Sentinel-2 reflectance images and the spectral information of MODIS reflectance images. The combination uses the maximum a posteriori estimation by establishing statistical relationships between the standardized image and the original image reflectance image and between the standardized image and the MODIS image.

[0030] The implementation steps of the data accuracy evaluation after reflectivity normalization in S4 are as follows:

[0031] Compare and analyze the GF-1, GF-6 and Sentinel-2 satellite images with MODIS images before and after reflectance standardization;

[0032] The calculation formula for quantitative analysis is:

[0033]

[0034]

[0035]

[0036] Among them, SS res is the residual sum of squares, yi is the ith observation, is the predicted value for the ith observation, SS tot is the total sum of squares, is the average of all observations and m is the number of predictions.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention uses MODIS data as a reference data set, conducts reflectance standardization research on GF-1, GF-6 and Sentinel-2 satellites, uses PSF to simulate the characteristics and spectral response function of MODIS sensors, models PSF as a Gaussian distribution function, and obtains the optimal standard deviation; degrades GF-1, GF-6 and Sentinel-2 reflectance images to the spatial resolution of MODIS, establishes a correlation, and then applies a Bayesian framework to establish statistical relationships between standardized images and original image reflectance images, as well as between standardized images and MODIS images, thereby utilizing maximum a posteriori estimation.

[0039] The reflectivity-normalized GF-1, GF-6, and Sentinel-2 satellite reflectivity is equivalent to the data acquired by the virtual sensor in the nadir view with a spatial resolution of 15 m and the spectral response function characteristics of MODIS.

[0040] After the reflectivity of multi-source remote sensing satellites is standardized, the density of time series images is significantly enhanced, and the details of the vegetation index time series curve are smoothed and increased, which can make the monitoring of crops more refined and provide support for precision agriculture, thereby optimizing the use of agricultural resources and improving agricultural production efficiency and sustainability. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the overall process of the present invention.

[0042] Figure 2 Schematic diagram of true color visual inspection results on different types of satellites according to the present invention.

[0043] Figure 3 This is a schematic diagram of the normalized front reflectivity of the blue band on GF-1 according to the present invention.

[0044] Figure 4 This is a schematic diagram of the reflectivity after blue band standardization on GF-1 according to the present invention.

[0045] Figure 5 This is a schematic diagram of the standardized front reflectivity of the green band on GF-1 according to the present invention.

[0046] Figure 6 This is a schematic diagram of the reflectivity after normalization in the green band on GF-1 according to the present invention.

[0047] Figure 7 This is a schematic diagram of the normalized front reflectivity of the red band on GF-1 according to the present invention.

[0048] Figure 8 This is a schematic diagram of the reflectivity after red band standardization on GF-1 according to the present invention.

[0049] Fig. 9 This is a schematic diagram of the normalized front reflectivity in the near-infrared band on GF-1 according to the present invention.

[0050] Fig.10 This is a schematic diagram of the reflectivity of GF-1 after normalization in the near-infrared band according to the present invention.

[0051] Fig.11 This is a schematic diagram of the blue band normalized front reflectivity on GF-6 according to the present invention.

[0052] Fig.12 This is a schematic diagram of the reflectivity after blue band standardization on GF-6 according to the present invention.

[0053] Fig.13 This is a schematic diagram of the standardized front reflectivity of the green band on GF-6 according to the present invention.

[0054] Fig.14 This is a schematic diagram of the reflectivity after normalization in the green band on GF-6 according to the present invention.

[0055] Fig.15 This is a schematic diagram of the normalized front reflectivity of the red band on GF-6 according to the present invention.

[0056] Fig.16 This is a schematic diagram of the reflectivity after red band standardization on GF-6 according to the present invention.

[0057] Fig.17 This is a schematic diagram of the normalized front reflectivity in the near-infrared band on GF-6 according to the present invention.

[0058] Fig.18 This is a schematic diagram of the reflectivity of GF-6 after normalization in the near-infrared band according to the present invention.

[0059] Fig.19 This is a schematic diagram of the normalized reflectivity of the blue band on Sentinel-2 according to the present invention.

[0060] Fig. 20 This is a schematic diagram of the reflectivity after blue band standardization on Sentinel-2 according to the present invention.

[0061] Fig.21 This is a schematic diagram of the normalized front reflectivity of the green band on Sentinel-2 according to the present invention.

[0062] Fig. 22 This is a schematic diagram of the reflectivity after normalization in the green band on Sentinel-2 according to the present invention.

[0063] Fig.23 This is a schematic diagram of the normalized front reflectivity of the red band on Sentinel-2 according to the present invention.

[0064] Fig.24This is a schematic diagram of the reflectivity after red band standardization on Sentinel-2 according to the present invention.

[0065] Fig.25 This is a schematic diagram of the normalized front reflectivity in the near-infrared band on Sentinel-2 according to the present invention.

[0066] Fig.26 This is a schematic diagram of the reflectivity after normalization in the near-infrared band on Sentinel-2 according to the present invention.

[0067] Fig. 27 It is a schematic diagram of the time series comparison of the blue band reflectivity before and after GF-1, GF-6 and Sentinel-2 standardization of the present invention.

[0068] Fig.28 It is a schematic diagram of the time series comparison of the green band reflectance before and after GF-1, GF-6 and Sentinel-2 standardization of the present invention.

[0069] Fig.29 This is a schematic diagram of the time series comparison of the red band reflectance before and after GF-1, GF-6 and Sentinel-2 standardization in the present invention.

[0070] Fig.30 This is a schematic diagram of the time series comparison of the near-infrared band reflectivity before and after GF-1, GF-6 and Sentinel-2 standardization in the present invention.

[0071] Fig.31 This is a schematic diagram of the comparison of NDVI time series established by the present invention before and after GF-1, GF-6 and Sentinel-2 standardization. DETAILED DESCRIPTION

[0072] See also Figures 1 to 31 , the present invention is further described in conjunction with the accompanying drawings:

[0073] The study area selected by the present invention is located at the junction of Hengshui City and Xingtai City in Hebei Province, China. The study area is located in the North China Plain and belongs to the continental monsoon climate zone. It is warm and semi-arid, with rich agricultural climate resources. The soil in this area is loose and fertile, and the river channels are abundant. It has good arable land conditions, so the main land use type is arable land, and the main crop types are winter wheat and corn. The selected area is the 45×45 pixel range of MODIS. Since the effective resolution of MODIS's MCD43A4 is 463.3127m, and the GF-1, GF-6 and Sentinel-2 images are resampled to 15.4438m, the spatial resolution differs by 30 times, that is, GF-1, GF-6 and Sentinel-2 image data in the range of 1350×1350 pixels, and the image date is selected from January 1, 2021 to December 31, 2021.

[0074] like Figure 1 As shown, the method for standardizing the reflectivity of multi-source remote sensing satellites includes the following steps:

[0075] S1. Obtain GF-1, GF-6 and Sentinel-2 satellite data, as well as MODIS data, and perform radiometric calibration, orthorectification, geometric registration and other preprocessing on GF-1 and GF-6 data.

[0076] S2. Taking MODIS as the reference data set, PSF is used to simulate the imaging effect and spectral response function of MODIS, and the GF-1, GF-6 and Sentinel-2 data are degraded to the spatial resolution of MODIS to solve the problem of large difference in spatial resolution.

[0077] S3. Based on the Bayesian framework, the statistical relationship between the standardized image and the original image reflectance image, as well as between the standardized image and the MODIS image, is established, so as to obtain the reflectance standardized image using the maximum a posteriori estimation, that is, the data with a spatial resolution of 15 m and the spectral response function characteristics of MODIS.

[0078] S4. In order to verify the accuracy of the reflectivity of multi-source remote sensing satellite images before and after standardization and MODIS reflectivity, a combination of qualitative and quantitative analysis is used. Qualitative analysis chooses to visually inspect the true color difference of the image. In quantitative analysis, R2, RMSE and rRMSE are introduced for evaluation and analysis. The introduction of rRMSE allows the RMSE between different data sets to be compared and a more accurate evaluation can be obtained. The above is an evaluation based on a single remote sensing image. It is necessary to perform a time series analysis of the images before and after the satellite reflectivity is standardized, and at the same time explore the role of using multi-source remote sensing satellites in increasing the time series.

[0079] In one embodiment of the present invention, reflectivity standardization research is conducted on GF-1, GF-6 and Sentinel-2 data to explore the collaborative application of multi-source remote sensing data. Images of the 2021 time area are selected: 16 GF-1 images, 14 GF-6 images, and 35 Sentinel-2 images. MODIS data is used as a reference data set to conduct reflectivity standardization research. For Sentinel-2 data, the L2A level is downloaded, which is an orthophoto with precise geometric correction and does not require preprocessing. MODIS's MCD43A4 product is a verified level 3 product and can be used directly in this study. Therefore, only GF-1 and GF-6 need to be preprocessed. GF-1 and GF-6 are preprocessed by ENVI, mainly including radiation calibration, atmospheric correction and orthorectification. Note that in order to better match the high-resolution data with the Sentinel-2 data during geometric registration, the high-resolution data is output to a resolution of 15m during orthorectification. In this invention, the near-infrared band of the same phase Sentinel-2 is used as the reference image, and the automatic registration and orthophoto package (AROP) is used to batch geometrically register GF-1 and GF-6, with a registration accuracy within 0.5 pixels. So far, GF-1, GF-6 and Sentinel-2 have accurate geographic matching, laying the foundation for subsequent reflectivity standardization.

[0080] In step S2, the point spread function is modeled as a Gaussian function, and the parameters are empirically estimated by comparing the spatially degraded GF-1, GF-6, and Sentinel-2 reflectances and MODIS images, and the best standard deviation is defined by the highest R2. By comparing the R2 between the MODIS image and the spatially degraded GF-1, GF-6, and Sentinel-2 blue, green, red, and near-red band reflectance images, it can be found that R2 increases first and then decreases with the increase of the standard deviation value. There is an R2 peak, and the maximum R2 range is 0.51-0.94. Based on this, the best standard deviations for each band of GF-1, GF-6, and Sentinel-2 can be obtained, as shown in Table 1 below, ranging from 230m to 320m.

[0081] Table 1: R2(m) of the best PSF standard deviation values ​​in the four bands of MODIS

[0082] type Blue Green Red NIR GF-1 280 290 280 260 GF-6 250 230 240 270 Sentinel-2 300 300 310 320

[0083] In step S3, according to the Bayesian framework, the reflectance-standardized image is obtained, that is, data with a spatial resolution of 15 m and the spectral response function characteristics of MODIS. In order to verify the accuracy of the reflectance of the multi-source remote sensing satellite image before and after standardization and the MODIS reflectance, in step S4, a combination of qualitative analysis and quantitative analysis is used. The qualitative analysis selects the true color difference of the visual inspection image. In the visual inspection, MODIS images, as well as GF-1, GF-6 and Sentinel-2 images before and after standardization are displayed, that is, Figure 2 shown.

[0084] In the quantitative evaluation, the images of the blue, green, red and near-infrared band reflectance of GF-1, GF-6 and Sentinel-2 data before and after standardization were spatially degraded to achieve the same spatial resolution as MODIS, and compared with the MODIS reflectance of the corresponding bands. R2, RMSE and rRMSE were used for evaluation and analysis, as shown in Tables 2 to 4 below, to evaluate the accuracy of GF-1, GF-6 and Sentinel-2 reflectance after standardization, and then a scatter plot was used for a more intuitive display, as shown in Tables 2 to 4 below. Figure 3 to Figure 26 shown.

[0085] Table 2: Changes in the correlation between GF-1 reflectance and MODSI before and after normalization

[0086]

[0087] In Table 2 above, the rRMSE and R2 are improved the most in the near-infrared band, and the smallest improvement is achieved in the red band. This is related to the intersection rate of the spectral response functions of the GF-1WFV and MODSI sensors, which has the lowest intersection rate in the near-infrared band. At the same time, after reflectance normalization, the red band and the near-infrared band have the largest and smallest rRMSE.

[0088] Table 3: Changes in the correlation between GF-6 reflectance and MODSI before and after normalization

[0089]

[0090] Table 3 shows that the improved rRMSE and R2 before and after the GF-6 data is normalized with the MODIS reflectance are compared with the parameters of the original reflectance. The improvements in the blue, green, red and near-infrared bands are 28.1% and 0.12, 24.2% and 0.08, 21.6% and 0.06, and 27.9% and 0.08, respectively. It can be found that in the blue band, rRMSE and R2 have the greatest improvement, and in the red band, the smallest improvement.

[0091] Table 4: Changes in correlation between Sentinel-2 reflectivity and MODSI before and after normalization

[0092]

[0093] In Table 4 above, the rRMSE and R2 of Sentinel-2 data before and after normalization with MODIS reflectance are compared with the parameters of the original reflectance. The improvements in the blue, green, red and near-infrared bands are 1.6% and 0.26, 1.2% and 0.16, 0.6% and 0.04, and 0.8% and 0.09, respectively. It can be found that in the blue band, rRMSE and R2 have the greatest improvement, and in the red band it is the smallest improvement. At the same time, after reflectance normalization, the blue band and the red band have the largest and smallest rRMSE, which is consistent with the crossover rate of the spectral response function of the Sentinel-2 and MODIS sensors.

[0094] Through visual inspection and quantitative evaluation of the results after satellite reflectivity standardization, it was found that after standardization, the true color presentation of GF-1, GF-6 and Sentinel-2 images is closer to the reference data set MODIS data, and the correlation between the blue, green, red and near-red bands and the corresponding band reflectivity of MODIS is also significantly improved. However, the above evaluation is based on a single remote sensing image, and it is necessary to conduct a time series analysis of the images before and after the satellite reflectivity standardization, and at the same time explore the role of using multi-source remote sensing satellites in increasing the time series.

[0095] A crop planting point in the study area was selected to perform time series analysis on the reflectance of the blue, green, red and near-red bands of GF-1, GF-6 and Sentinel-2 images before and after the reflectance standardization, and a time series curve was generated. By comparing the time series curves of these different bands, it can be found that the reflectance of the original GF-1, GF-6 and Sentinel-2 images lacks consistency in the time series, and the connection points of different data sources are quite different, making it impossible to perform a systematic time series analysis. After standardizing the reflectance of multiple satellite sources using MODIS as the reference data set, the consistency of the data is significantly improved, and the time series curve is smoother.

[0096] like Figure 27 to Figure 30 As shown in the figure, compared with the HLS band reflectance time series, it also has similar smoothness and consistency. At the same time, the image density of time series images in the crop growing season increases significantly, which plays an important role in establishing the vegetation index time series set.

[0097] In order to explore the advantages of GF-1, GF-6 and Sentinel-2 images after reflectance standardization in establishing a time series vegetation index dataset, the NDVI time series of the crop planting point was analyzed. There are 16 GF-1 images, 14 GF-6 images, and 35 Sentinel-2 images.

[0098] First, the NDVI of each satellite before and after reflectivity standardization is compared. The number of GF-1 and GF-6 images is too small, and they can only show the growth period of crops to a certain extent, and cannot monitor the growth of crops in detail. The number of images of Sentinel-2 data is large, but the original NDVI curve is not smooth enough due to large fluctuations. After reflectivity standardization, the consistency and smoothness of the NDVI time series curve are enhanced. A unified NDVI time series curve is established for GF-1, GF-6 and Sentinel-2 images. It can be found that when reflectivity standardization is not performed, the consistency of NDVI data between satellites is poor and the fluctuation is large, which is not conducive to the establishment of the NDVI time series, and the monitoring of crops may also be biased. After reflectivity standardization, the fluctuation in the curve is significantly reduced, such as Fig.31 As shown in the figure, the circled position in the box shows similar characteristics compared with the NDVI time series generated with HLS as the data source. Using MODIS data as a reference significantly eliminates the abnormal fluctuation value in the NDVI curve, which is of great significance for establishing a time series vegetation index.

[0099] The above are only preferred embodiments of the present invention, which are only used to help understand the method and core idea of ​​the present application. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the idea of ​​the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

[0100] The present invention solves the problem in the prior art that in the coordinated use of multi-source remote sensing satellites, due to different spectral responses, band configurations and detection characteristics between different satellite data sensors, the data obtained by different sensors have differences in reflectivity and cannot form a unified standard, thereby making the data inconsistent and affecting the precision and accuracy of the data. By conducting reflectivity standardization research on different satellites, obtaining the best standard deviation of MODIS sensor data, establishing correlations between different satellites, establishing statistical relationships between standardized images, original images and MODIS images, and performing maximum a posteriori estimation, data with a spatial resolution of 15m and MODIS spectral response function characteristics are obtained, which greatly improves the density and details of the time series on the image, makes monitoring more refined, provides a reliable basis for monitoring the growth of crops, etc., promotes the scientific development of agriculture, and improves the optimal allocation of its production resources, efficient production and sustainable development.

Claims

1. A method for standardizing reflectivity of multi-source remote sensing satellites, characterized in that: The following steps are involved: S1, obtain GF-1, GF-6 and Sentinel-2 satellite data, as well as MODIS data, and pre-process GF-1 and GF-6 data; S2, taking MODIS data as a reference data set, using PSF to simulate the imaging effect and spectral response function of MODIS, and degenerating the data of GF-1, GF-6 and Sentinel-2 to the spatial resolution of MODIS, so as to solve the problem of large difference in spatial resolution; S3, based on the Bayesian framework, the statistical relationship between the standardized image and the original image reflectance image, as well as between the standardized image and the MODIS image, is established, so as to obtain the reflectance standardized image using the maximum a posteriori estimation, i.e., data with a spatial resolution of 15 m and the spectral response function characteristics of MODIS; S4, a combination of qualitative and quantitative analysis methods was used to verify the accuracy of the reflectance of multi-source remote sensing satellite images based on a single remote sensing image before and after normalization and MODIS reflectance; Among them, qualitative analysis selects visual inspection of the true color difference of the image; quantitative analysis introduces the determination coefficient R 2 , root mean square error RMSE and relative root mean square error rRMSE are used for evaluation and analysis, where rRMSE is introduced to compare the RMSE between different data sets to obtain a more accurate evaluation; Subsequently, the satellite reflectivity images before and after standardization were analyzed in time series, and the verified reflectivity standardization product HLS data was introduced to conduct time series comparative analysis with the images before and after standardization to prove the feasibility of this method; Harmonized Landsat Sentinel-2 is HLS data.

2. The method for standardizing the reflectivity of multi-source remote sensing satellites according to claim 1, characterized in that: The specific process of preprocessing in S1 is as follows: Download GF-1 and GF-6 images of the experimental area from the website of China Resources Satellite Application Center, download L2A-level products of Sentinel-2 data of the experimental area from the official website of ESA, and download MODIS MCD43A4 data and Harmonized Landsat Sentinel-2 (HLS) data from the official website of NASA; perform image preprocessing on GF-1 and GF-6 images, including radiation calibration, atmospheric correction, orthorectification and geometric registration, and use Sentinel-2 images as the reference image for geometric registration.

3. The method for standardizing the reflectivity of multi-source remote sensing satellites according to claim 1, characterized in that: The specific steps of S2 are: The point spread function is modeled as a Gaussian distribution curve to characterize the spectral response function. The reflectance data of GF-1, GF-6 and Sentinel-2 should be as similar as possible to MODIS when downgraded to MODIS resolution by considering the point spread function of the MODIS sensor. The formula is: A 463m (λ)=subsamplepsf(λ)*A 15m (l); in: A 463m (λ) is the degraded GF-1, GF-6 and Sentinel-2 images at the MODIS 463.3127m resolution with a wavelength of λ; A 15m (λ) is the preprocessed 15.4438m resolution reflectance image of GF-1, GF-6 and Sentinel-2 at wavelength λ; PSF(λ) is the n×n convolution kernel matrix representing MODIS PSF, * represents convolution, and subsample is the subsampling of every 30 pixels of the convolved image, where 30≈463.3127 / 15.4438; The effective spatial resolution of MODIS images is about 600m to 1000m. Set n to 90, that is, use 90×90 pixels of high-resolution images for convolution, and model the PSF as a Gaussian function: where psfλ(i,j) is the filter weight with i and j∈(1,2,…,90), so i=46,j=46 is the center of the 90×90 filter; σ λ is the standard deviation specific to the study area and band; σλ is empirically derived by maximizing the correlation coefficient between degraded GF-1, GF-6, Sentinel-2, and MODIS images, as follows: R 2 For two variables A 463m (λ,σ λ ) and M 463m The coefficient of determination of (λ) is 463m (λ,σ λ ) as σ λ The function of is derived through the point spread function formula, M 463m (λ) represents the MODIS NBAR image with a resolution of 463.3127 m.

4. The method for standardizing the reflectivity of multi-source remote sensing satellites according to claim 3, characterized in that: The λ is the blue, green, red and near infrared bands.

5. The method for standardizing the reflectivity of multi-source remote sensing satellites according to claim 1, characterized in that: The specific steps of the Bayesian method in S3 are: The point spread function is introduced to represent the prior distribution. The reflectance of the image after reflectance standardization is obtained by combining the spatial information of GF-1, GF-6 and Sentinel-2 reflectance images and the spectral information of MODIS reflectance images. The combination of the spatial information of the reflectance image and the spectral information of the reflectance image is used to utilize the maximum a posteriori estimation by establishing statistical relationships between the standardized image and the original image reflectance image and between the standardized image and the MODIS image.

6. The method for standardizing the reflectivity of multi-source remote sensing satellites according to claim 1, characterized in that: The implementation steps of the data accuracy evaluation after reflectivity standardization in S4 are as follows: The GF-1, GF-6 and Sentinel-2 satellite images before and after reflectance normalization were selected for comparative analysis with MODIS images; The calculation formula for quantitative analysis is: Among them, SS res is the residual sum of squares, yi is the ith observation, is the predicted value for the ith observation, SS tot is the total sum of squares, is the average of all observations and m is the number of predictions.

Images