A method for atmospheric correction of remote sensing data

By adopting efficient screening and spatiotemporal expansion strategies in remote sensing data processing, the problem of insufficient accuracy and efficiency in the atmospheric correction method of remote sensing data is solved, efficient AOD inversion and atmospheric correction are achieved, and data quality is improved.

CN116797772BActive Publication Date: 2024-11-22BEIJING DATA INTELLIGENCE INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310744753.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-06-21
Publication Date
2024-11-22
Estimated Expiration
2043-06-21

AI Technical Summary

Technical Problem

The existing atmospheric correction methods for remote sensing data are insufficient in terms of accuracy and efficiency, especially the methods based on the statistical information of the remote sensing image itself have poor consistency and low accuracy, and require a large amount of manpower and material resources to perform AOD inversion.

Method used

Representative data is selected by setting an efficient screening method, adaptive spatiotemporal expansion strategies are used to simulate and calculate AOD, and atmospheric correction is used to reduce the cost of data inversion and manpower and material investment. The specific steps include obtaining remote sensing data, calculating normalized vegetation index, screening data, calculating AOD using the atmospheric radiation transmission model, and obtaining the AOD of the entire set of data through the spatiotemporal expansion strategy, and finally performing atmospheric correction.

Benefits of technology

It improves the accuracy of AOD inversion, reduces data processing costs, achieves efficient atmospheric correction effects, and enhances the quality and representativeness of remote sensing data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116797772B_ABST
    Figure CN116797772B_ABST
Patent Text Reader

Abstract

The present invention provides a remote sensing data atmospheric correction method, which relates to the field of data correction, including: S1 obtaining a group of remote sensing data and corresponding metadata; S2 calculating the normalized vegetation index of the remote sensing data to obtain a first classification set, a second classification set, and a third classification set; S3 screening the first classification set, the second classification set, and the third classification set according to the metadata to obtain a first subset, a second subset, and a third subset; S4 calculating the AOD of the first subset, the second subset, and the third subset according to the metadata and the atmospheric radiation transmission model; S5 using a time-space expansion strategy to expand the AOD of the first subset, the second subset, and the third subset to obtain the AOD of the entire group of remote sensing data; S6 performing a unified atmospheric correction on the remote sensing data according to the AOD to obtain the surface reflectivity of the remote sensing data. The method provided by the present invention can greatly improve the accuracy of AOD inversion, and does not require a large amount of data for inversion, thereby reducing costs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of data correction, and in particular to a remote sensing data atmospheric correction method. Background Art

[0002] Aerosols in the atmosphere are intermediary substances between ground and satellite sensors, and inevitably affect the data obtained by satellite remote sensing. As the aerosol increases, the quality of the image obtained will also be affected. In order to make full use of the surface information obtained by satellite remote sensing, we need to correct the information caused by the atmosphere in the image. Since the 1970s, scientists have conducted a lot of research on atmospheric correction of remote sensing data. At present, atmospheric correction of remote sensing data is mainly divided into two categories: atmospheric correction methods based on the statistical information of remote sensing images themselves and atmospheric correction methods based on specific surface features; the former method is simple to implement and efficient, but has poor consistency and low accuracy; the latter method requires the inversion of aerosol and atmospheric water vapor information in the atmosphere, and needs to solve the atmospheric radiation transmission model. The correction effect is related to the inversion accuracy of atmospheric aerosol and atmospheric water vapor, and the efficiency is low.

[0003] The atmospheric correction method based on the statistical information of remote sensing images uses the difference between the histograms of remote sensing images in clean areas (areas containing only background aerosols or a small amount of aerosols) and non-clean areas (areas containing aerosols) to perform statistical analysis and correct the images in non-clean areas. This type of method mainly includes histogram matching method, cluster matching method, tasseled cap transformation method, smoke optimization transformation method, etc. Among them, the smoke optimization transformation method has been integrated into the ATCOR software and has been widely used. However, this type of method first requires that the image has multiple similar surfaces, and these similar surfaces have a relatively clean atmosphere. If there is no clean area, atmospheric correction cannot be performed; at the same time, the clean area needs to be selected manually and does not have automation capabilities.

[0004] The accurate correction of atmospheric effects depends on the accurate inversion of AOD. However, remote sensing images in the visible and near-infrared bands are highly coupled information between the ground and the atmosphere. To accurately obtain surface information or atmospheric information, one must first know one of the information and then use the radiation transfer equation to solve the other information. The accuracy of atmospheric correction of remote sensing images and the degree of knowledge of surface information determine the accuracy of atmospheric correction. Domestic and foreign research on aerosol remote sensing has made various assumptions about the surface conditions and inverted aerosol information for atmospheric correction. However, the amount of remote sensing data is huge, and it will take a lot of manpower and material resources to perform atmospheric correction after AOD inversion on all remote sensing data. Summary of the invention

[0005] Based on the above technical problems, the present invention provides a method for atmospheric correction of remote sensing data, which selects an appropriate amount of representative data by setting an efficient and reasonable screening method, and simulates and calculates AOD through an adaptive time and space expansion strategy, thereby obtaining the AOD of the entire set of remote sensing data, and uses AOD for atmospheric correction. The method provided by the present invention can greatly improve the accuracy of AOD inversion, and does not require a large amount of data for inversion, thereby reducing costs.

[0006] In order to achieve the above technical purpose, the present invention provides a remote sensing data atmospheric correction method comprising:

[0007] S1 obtains a set of remote sensing data and corresponding metadata, and preprocesses the remote sensing data;

[0008] S2 calculates the normalized vegetation index of the remote sensing data, sets two thresholds, and classifies the remote sensing data according to the normalized vegetation index based on the two thresholds to obtain three classification sets, namely, the first classification set, the second classification set, and the third classification set;

[0009] S3 screens the first classification set, the second classification set, and the third classification set respectively according to the metadata to obtain a first subset, a second subset, and a third subset;

[0010] S4 calculates the AOD of the first subset, the second subset, and the third subset respectively according to the metadata and the atmospheric radiation transfer model;

[0011] S5 uses a spatiotemporal expansion strategy to expand the AODs of the first subset, the second subset, and the third subset respectively to obtain an expanded AOD, and obtains the AOD of the entire set of remote sensing data according to the expanded AOD and the AODs of the first subset, the second subset, and the third subset;

[0012] S6 uses the atmospheric radiation transfer model to establish an atmospheric correction lookup table, and performs a unified atmospheric correction on the remote sensing data according to the AOD based on the atmospheric correction lookup table to obtain the surface reflectivity of the remote sensing data.

[0013] In a specific embodiment of the present invention, step S1 includes:

[0014] S11 obtains the band value of each pixel of each remote sensing data by calculation;

[0015] S12 obtains historical remote sensing data, uses the long-term historical remote sensing data to construct a clear sky background field containing a blue light band, and calculates a cloud detection threshold based on the clear sky background field containing a blue light band and the blue light band reflectance of each pixel of the remote sensing data;

[0016] S13 uses the cloud detection threshold to perform cloud detection on the remote sensing data, removes cloud and cloud shadow pixels, and obtains pre-processed remote sensing data.

[0017] In a specific embodiment of the present invention, step S2 includes:

[0018] S21 calculates the normalized vegetation index of remote sensing data. The calculation formula is as follows:

[0019]

[0020] Among them, NDVI is the normalized vegetation index, the NDVI value range is [-1,1], ρ NIR is the near infrared band value, ρ RED is the reflectivity of red light band;

[0021] S22 sets a first threshold value ρ1 and a second threshold value ρ2, the first threshold value ρ1 is less than the second threshold value ρ2, and the first threshold value ρ1 and the second threshold value ρ2 divide the value range of NDVI [-1,1] into three value segments, namely, a first value segment [-1,ρ1), a second value segment [ρ1,ρ2], and a third value segment [ρ2,1];

[0022] S23 compares the NDVI value of remote sensing data with three value ranges:

[0023] If the NDVI value of a remote sensing data falls within the range of the first value segment, the remote sensing data is classified into the first classification set;

[0024] If the NDVI value of a remote sensing data falls within the range of the second value segment, the remote sensing data is classified into the second classification set;

[0025] If the NDVI value of a remote sensing data falls within the range of the third value segment, the remote sensing data is classified into the third classification set.

[0026] In a specific embodiment of the present invention, step S3 includes:

[0027] S31 extracts the data acquisition range and data acquisition date from the metadata, and performs quality assessment on the remote sensing data according to the cloud detection results to obtain the quality score of the remote sensing data;

[0028] S32 arranges the remote sensing data in chronological order from recent to far according to the data acquisition date, and divides them into sections with months as nodes to obtain multiple sets of section sets;

[0029] S33 divides the remote sensing data according to the data acquisition range according to the preset divisions, and obtains multiple groups of range sets based on the coverage of the preset divisions reaching 60%;

[0030] S34 sorts each group of segment sets and each group of range sets from high to low according to the quality scores, and cross-screens the sorted segment sets and the sorted range sets to obtain a screening set;

[0031] S35 compares the screening set with the first classification set, the second classification set, and the third classification set:

[0032] If the data in the screening set belongs to the first classification set, the data is classified into the first subset;

[0033] If the data in the screening set belongs to the second classification set, the data is classified into the second subset;

[0034] If the data in the screening set belongs to the third classification set, the data is classified into the third subset.

[0035] In a specific embodiment of the present invention, step S34 includes:

[0036] The sorted segment set is denoted as {{A i}} m , and record the sorted range set as {{B j}} n , where m is the number of segments, A i is the i-th segment data of the current group, n is the number of groups in the range set, B j is the j-th range data of the current group;

[0037] Set m groups {A i The segment data with the top 50% quality scores in} are taken as the segment set;

[0038] Set n groups {B j The range data with the top 30% quality scores in} are taken as the range set;

[0039] The segment set and range set are merged and deduplicated to obtain the filter set.

[0040] In a specific embodiment of the present invention, step S4 includes:

[0041] S41 extracts calibration information and solar geometry information from metadata;

[0042] S42 calculates the apparent reflectance of each data in the first subset, the second subset, and the third subset respectively according to the calibration information and the solar geometry information;

[0043] S43 uses the atmospheric radiation transfer model to establish an AOD inversion lookup table;

[0044] S44 calculates the AOD of the first subset, the second subset, and the third subset respectively according to the metadata, the AOD inversion lookup table, and the apparent reflectivity.

[0045] In a specific embodiment of the present invention, step S5 includes:

[0046] S51 time-space expansion strategy includes time expansion strategy and space expansion strategy;

[0047] S52 performs time series expansion on the AODs of the first subset, the second subset, and the third subset respectively according to the time series expansion strategy to obtain a first time series simulation AOD, a second time series simulation AOD, and a third time series simulation AOD;

[0048] S53 spatially expands the AODs of the first subset, the second subset, and the third subset according to the spatial expansion strategy to obtain a first spatial simulation AOD, a second spatial simulation AOD, and a third spatial simulation AOD;

[0049] S54 calculates the extended AOD in a weighted manner:

[0050] Calculate the AOD of the first subset, the first temporal simulation AOD, and the first spatial simulation AOD according to a first weighting method to obtain a first extended AOD;

[0051] The AOD of the second subset, the second temporal simulation AOD, and the second spatial simulation AOD are calculated according to a second weighting method to obtain a second extended AOD;

[0052] The AOD of the third subset, the third time series simulation AOD, and the third space simulation AOD are calculated according to a third weighting method to obtain a third extended AOD;

[0053] S55 combines the first extended AOD, the second extended AOD, the third extended AOD, the AOD of the first subset, the AOD of the second subset, and the AOD of the third subset as the AOD of the remote sensing data.

[0054] In a specific embodiment of the present invention, the timing expansion strategy is:

[0055] Step 1: Establish three time segment sequences according to the data acquisition time in the three classification sets. Each time segment is one month, and the sequence is arranged from recent to far.

[0056] Step 2: According to the data acquisition time of each data in the three subsets, a distribution diagram of the landing point of the subset in the corresponding time segment sequence is formed to obtain three distribution diagrams of the landing point;

[0057] Step 3, calculate the number of landing points in each time segment in a single landing point distribution diagram, and use the data corresponding to the time segments with the last 40% of the number of landing points as the time data set to be expanded for the corresponding subset;

[0058] Step 4: Taking the collection time of a single time data to be expanded of the time data set to be expanded as the intermediate node, selecting data within a week centered on the intermediate node in the corresponding classification set, and eliminating data with a quality score lower than 50%, to obtain a set of time candidate data sets;

[0059] Step 5: Repeat step 4 to obtain multiple sets of candidate time data sets for each time data set to be expanded;

[0060] Step 6: Calculate the AOD of all the time candidate data in a set of time candidate data sets, and take the average of all the AODs in the set of time candidate data sets and the AOD of the corresponding time data to be expanded as the candidate time series simulation AOD;

[0061] Step 7: Repeat step 6, and obtain multiple candidate time series simulation AODs for each subset. The multiple candidate time series simulation AODs are classified according to the affiliation of the subsets to obtain the first time series simulation AOD, the second time series simulation AOD, and the third time series simulation AOD.

[0062] In a specific embodiment of the present invention, the space expansion strategy is:

[0063] The first step is to use the merged area of ​​the data acquisition range in each set of the first classification set, the second classification set, and the third classification set as their respective target areas, calculate the union of the data acquisition range of each subset, and use the mean of the AOD of all the data in the subset as the AOD of the corresponding union;

[0064] The second step is to compare the single union with the corresponding target area, and take the non-intersecting area as the area to be expanded of the corresponding subset;

[0065] Step 3: spatially interpolate the AOD of the single union to the corresponding area to be expanded with a step size of 3% until the area to be expanded of the corresponding subset is completely covered to obtain the AOD of the single area to be expanded;

[0066] Step 4: Repeat steps 2 and 3 to obtain AODs of three areas to be expanded, which are used as the first space simulation AOD, the second space simulation AOD, and the third space simulation AOD respectively.

[0067] In a specific embodiment of the present invention:

[0068] The first weighting method is expressed as:

[0069]

[0070] stf1∈[1,F1],i1∈[1,n1],j1∈[1,m1],k1∈[1,h1],α1,β1,a,b∈[0,1]

[0071] in, is the f1th first extended AOD, F1 is the number of first extended AODs, α1 and β1 are weight coefficients, a and b are adjustment coefficients, n1 is the number of data in the first subset, m1 is the number of first time series simulation AODs, h1 is the number of first space simulation AODs, represents the AOD of the i1th data in the first subset, represents the j1th first time series simulation AOD, represents the k1th first spatial simulation AOD;

[0072] The second weighting method is expressed as:

[0073]

[0074] stf2∈[1,F2],i2∈[1,n2],j2∈[1,m2],k2∈[1,h2],α2,β2∈[0,1]

[0075] in, is the f2th second extended AOD, F2 is the number of second extended AODs, α2 and β2 are weight coefficients, n2 is the number of data in the second subset, m2 is the number of second temporal simulation AODs, h2 is the number of second spatial simulation AODs, represents the AOD of the i2th data in the second subset, represents the j2th second time series simulation AOD, represents the k2th second space simulation AOD;

[0076] The third weighting method is:

[0077]

[0078] stf3∈[1,F3],i3∈[1,n3],j3∈[1,m3],k3∈[1,h3],α3,β3∈[0,1]

[0079] in, is the f3th third extended AOD, F3 is the number of third extended AODs, α3 and β3 are weight coefficients, n3 is the number of data in the third subset, m3 is the number of third time series simulation AODs, and h3 is the number of third space simulation AODs. represents the AOD of the i3th data in the third subset, represents the j3rd time series simulation AOD, represents the k3th third space simulation AOD.

[0080] The beneficial effects of the present invention include:

[0081] (1) The present invention designs three weighting methods and comprehensively considers the characteristics of different scenes. The calculated extended AOD can be more consistent with the original AOD of remote sensing data, improve the accuracy of simulation, and increase the accuracy of inversion. The subsequent atmospheric correction can achieve the effect of fine correction.

[0082] (2) The screening set obtained from the two levels of time and scope in the present invention has high quality. After the screening set is divided according to affiliation, the first subset, the second subset, and the third subset obtained will not differ too much in quantity and will be more representative. It can reasonably reflect the scenarios covered by the first classification set, the second classification set, and the third classification set respectively. When performing subsequent simulation inversion, using the three subsets for representative inversion can reduce errors and improve the accuracy of inversion.

[0083] (3) The present invention can expand the AOD of the three subsets from the time dimension through the time series expansion strategy. The expanded time series simulation AOD is used as the AOD estimation of other data in the time dimension. The time series simulation AOD has high accuracy.

[0084] (4) The present invention expands the AOD from two dimensions, time and space, and can simulate the AOD of large-scale data with small-scale data to the greatest extent, thereby reducing the cost of manpower and material resources, and taking into account the different characteristics of remote sensing data in time and space. The simulated AOD obtained can greatly represent the AOD of remote sensing data, thereby improving the accuracy of inversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0086] Figure 1 A schematic diagram of a method according to an embodiment of the present invention;

[0087] Figure 2 The technical flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. It should be noted that as long as there is no conflict, the various embodiments and various features of the embodiments in the present invention can be combined with each other, and the technical solutions formed are all within the protection scope of the present invention.

[0089] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0090] See also Figure 1 and Figure 2 The present invention provides a remote sensing data atmospheric correction method, comprising:

[0091] S1 obtains a set of remote sensing data and corresponding metadata, and preprocesses the remote sensing data.

[0092] Specifically, step S1 includes:

[0093] S11 obtains the band value of each pixel of each remote sensing data by calculation;

[0094] S12 obtains a set of historical remote sensing data, which is a long time series, and constructs a clear sky background field containing a blue light band by using the long time series of historical remote sensing data. According to the clear sky background field containing the blue light band and the blue light band reflectivity of each pixel of the remote sensing data, a cloud detection threshold is calculated;

[0095] S13 uses the cloud detection threshold to perform cloud detection on the remote sensing data, removes cloud and cloud shadow pixels, and obtains pre-processed remote sensing data.

[0096] Step S1 is described with a specific embodiment:

[0097] In the specific implementation, cloud pixels are identified for long-term historical remote sensing data, and cloud shadow pixels are identified based on the flooding algorithm. Cloud and cloud shadow pixels are eliminated, and clear sky pixels are screened out to obtain T clear sky images. For pixels at the same position in these T clear sky images, the minimum value of the blue light band reflectance is obtained, and all minimum values ​​are arranged in the corresponding pixel arrangement order to obtain a clear sky background field containing the blue light band.

[0098] The band values ​​of each pixel (x, y) of each remote sensing data are calculated based on the metadata, including: blue light band reflectance Band1(x, y), green light band reflectance Band2(x, y), red light band reflectance Band3(x, y), near infrared band reflectance Band4(x, y) and brightness temperature Band15(x, y) of the fifteenth band.

[0099] Get the surface information Band1 of the blue light band of each pixel in the clear sky background field clear-sky(x,y), the blue light band reflectance of each pixel (x, y) of remote sensing data is compared with Band1 clear-sky(x,y) The difference is obtained by subtraction, and the difference is subjected to multi-scale retinal enhancement with color rejuvenation to obtain the reflectance difference. The cloud detection threshold is extracted from the reflectance difference image by the Otsu algorithm.

[0100] According to the cloud detection threshold, it is judged whether each pixel in the remote sensing data is a cloud pixel, and the clouds are identified. At the same time, the flooding algorithm is used to identify cloud shadows, and the clouds and cloud shadows in the remote sensing data are identified and removed.

[0101] S2 calculates the normalized vegetation index of the remote sensing data, sets two thresholds, and classifies the remote sensing data according to the normalized vegetation index based on the two thresholds to obtain three classification sets, namely the first classification set, the second classification set, and the third classification set.

[0102] Specifically, step S2 includes:

[0103] S21 calculates the normalized vegetation index of remote sensing data. The calculation formula is as follows:

[0104]

[0105] Among them, NDVI is the normalized vegetation index, the NDVI value range is [-1,1], ρ NIR is the near infrared band value, ρ RED is the reflectivity in the red light band.

[0106] S32 sets the first threshold ρ1 and the second threshold ρ2. The first threshold ρ1 is smaller than the second threshold ρ2. The first threshold ρ1 and the second threshold ρ2 divide the value range of NDVI [-1,1] into three value segments, namely the first value segment [-1,ρ1), the second value segment [ρ1,ρ2), and the third value segment [ρ2,1].

[0107] S33 compares the NDVI value of remote sensing data with three value ranges:

[0108] If the NDVI value of a remote sensing data falls within the range of the first value segment, the remote sensing data is classified into the first classification set;

[0109] If the NDVI value of a remote sensing data falls within the range of the second value segment, the remote sensing data is classified into the second classification set;

[0110] If the NDVI value of a remote sensing data falls within the range of the third value segment, the remote sensing data is classified into the third classification set.

[0111] In this embodiment, the first threshold value ρ1 is 0, and the second threshold value ρ2 is 0.5. The present invention sets two thresholds to divide the remote sensing data according to the NDVI value, wherein the NDVI values ​​of the data in the first classification set are all less than 0, and the data in this classification set contains most of the complex surfaces, that is, the scenes in these data can be regarded as mostly urban construction land. The NDVI value of the data in the second classification set is 0-0.5, and the data in this classification set includes both urban construction land and vegetation areas, that is, the scenes in these data can be regarded as urban construction land and vegetation areas with similar proportions. The NDVI value of the data in the third classification set is 0.5-1, and the data in this classification set contains most of the vegetation areas, and the vegetation is dense, that is, the scenes in these data can be regarded as mostly vegetation areas.

[0112] S3 screens the first classification set, the second classification set, and the third classification set respectively according to the metadata to obtain a first subset, a second subset, and a third subset.

[0113] Specifically, step S3 includes:

[0114] S31 extracts the data acquisition range and data acquisition date from the metadata, and evaluates the quality of the remote sensing data according to the cloud detection results to obtain the quality score of the remote sensing data.

[0115] In one embodiment of the present invention, after removing cloud and cloud shadow pixels from remote sensing data, the quality of the remote sensing data is assessed based on the proportion of these removed pixels in the corresponding single remote sensing data, and the proportion is converted into a quality score of 0%-100%. In this embodiment, the proportion can be inverted according to 100% and the difference therebetween to set the corresponding quality score. For example, if the proportion of removed pixels in a remote sensing data is 16%, then its corresponding quality score is 84%.

[0116] S32 arranges the remote sensing data in chronological order from near to far according to the data acquisition date, and divides the data into sections with months as nodes to obtain multiple sets of section sets.

[0117] Specifically, the remote sensing data acquired in this embodiment is data within the past year, and after the data acquisition date is divided into months, a total of 12 segment sets are obtained.

[0118] S33 divides the remote sensing data according to the data acquisition range and the preset regions, and obtains multiple range sets based on the coverage of the preset regions reaching 60%.

[0119] Specifically, the remote sensing data acquired in this embodiment is data of the Central China region, which can be divided into provinces, cities, counties, districts, etc. according to administrative divisions. In this embodiment, the preset divisions are based on cities, and the remote sensing data with a coverage of 60% for a certain city is used as a set of range sets corresponding to the city. For example, taking Wuhan as the preset division, there are 5 images in the remote sensing data with a coverage of 60% for Wuhan, and these 5 images are used as a set of range sets.

[0120] S34 sorts each group of segment sets and each group of range sets from high to low according to the quality scores, and cross-screens the sorted segment sets and the sorted range sets to obtain a screened set.

[0121] Specifically, step S34 includes:

[0122] The sorted segment set is denoted as {{A i}} m , and record the sorted range set as {{B j}} n , where m is the number of segments, A i is the i-th segment data of the current group, n is the number of groups in the range set, B j is the j-th range data of the current group;

[0123] Set m groups {A i The segment data with the top 50% quality scores in} are taken as the segment set;

[0124] Set n groups {B j The range data with the top 30% quality scores in} are taken as the range set;

[0125] The segment set and range set are merged and deduplicated to obtain the filter set.

[0126] S35 compares the screening set with the first classification set, the second classification set, and the third classification set:

[0127] If the data in the screening set belongs to the first classification set, the data is classified into the first subset;

[0128] If the data in the screening set belongs to the second classification set, the data is classified into the second subset;

[0129] If the data in the screening set belongs to the third classification set, the data is classified into the third subset.

[0130] In this embodiment, the screening set obtained from the two levels of time and scope has a higher quality. After the screening set is divided according to affiliation, the first subset, the second subset, and the third subset obtained will not differ too much in quantity and will be more representative. It can reasonably reflect the scenarios covered by the first classification set, the second classification set, and the third classification set respectively. When performing subsequent simulation inversion, using three subsets for representative inversion can reduce errors and improve the accuracy of inversion.

[0131] S4 calculates the AOD of the first subset, the second subset, and the third subset respectively according to the metadata and the atmospheric radiation transfer model.

[0132] Specifically, step S4 includes:

[0133] S41 extracts calibration information and solar geometry information from metadata;

[0134] S42 calculates the apparent reflectance of each data in the first subset, the second subset, and the third subset respectively according to the calibration information and the solar geometry information;

[0135] S43 uses the atmospheric radiation transfer model to establish an AOD inversion lookup table;

[0136] S44 calculates the AOD of the first subset, the second subset, and the third subset respectively according to the metadata, the AOD inversion lookup table, and the apparent reflectivity.

[0137] In this embodiment, the method for calculating AOD in the first subset, the second subset, and the third subset is the same, and a unified description is given below based on a specific embodiment:

[0138] First, obtain the calibration information, including the maximum and minimum radiation brightness values, the maximum and minimum DN values ​​of the pixels, the state parameters of the sensor, and the solar geometry information, including the solar zenith angle and the solar azimuth angle.

[0139] The apparent reflectance is calculated based on the calibration announcement and solar geometry information. The calibration formula is:

[0140] L λ =(LMAX λ -LMIN λ ) / (Q calmax -Q calmin )×(Q cal -Q calmin )+LMIN λ

[0141] in:

[0142] L λ : Spectral radiance obtained by the sensor, unit is W / (m 2 sr.μm);

[0143] Q cal : DN value corresponding to the pixel;

[0144] Q calmax and Q calmin : are the maximum and minimum DN values ​​of pixels respectively;

[0145] LMAX λ and LMIN λ : are the spectral radiances of the maximum and minimum DN values ​​of the corresponding pixels, in W / (m 2 sr.μm).

[0146] The conversion formula from spectral radiance to apparent reflectance is:

[0147] ρ λ =π.L λ .d 2 / (ESUN λ .cosθ s )

[0148] in:

[0149] ρ λ : apparent reflectance, unitless;

[0150] π: Pi, approximately equal to 3.14159, no unit;

[0151] d: distance between the Earth and the Sun, astronomical unit;

[0152] ESUN λ : Average zenith solar irradiance, unit: W / (m 2 .μm);

[0153] θ s : Solar zenith angle, unit: degree.

[0154] Using the 6S atmospheric radiation transfer model, the AOD inversion lookup table is established, and its parameters are set as follows:

[0155] 16 surface reflectances: 0-0.15, step size 0.01;

[0156] 13 satellite zenith angles: 0-72, step size is 6, unit is degree;

[0157] 13 solar zenith angles: 0-72, step size is 6, unit is degree;

[0158] 19 relative azimuths: 0-180, step size is 10, unit is degree;

[0159] 14 AOD: values ​​are 0.0001, 0.1-1.0, 0.1-1.2, 0.1-1.5, 0.1-2.0, with a step size of 0.1;

[0160] 2 atmosphere types: mid-latitude summer and mid-latitude winter;

[0161] 1 aerosol type: continental aerosol;

[0162] Output parameters: atmospheric radiation reflectivity, total transmittance of incident light from the top of the atmosphere to the ground, and total transmittance upward into the field of view of the satellite sensor.

[0163] According to the apparent reflectivity calculated above and the three parameters output by the 6S atmospheric radiation transfer model, the surface reflectivity is calculated by the following formula:

[0164]

[0165] in:

[0166] θ s : solar zenith angle;

[0167] θ v : satellite zenith angle;

[0168] Relative azimuth;

[0169] Apparent reflectance;

[0170] Atmospheric radiation reflectivity;

[0171] T(θ s ): the total transmittance of incident light from the top of the atmosphere to the ground;

[0172] T(θ v ): total transmittance in the direction of upward into the field of view of the satellite sensor;

[0173] S: downward hemispheric reflectivity of the bottom layer of the atmosphere, which can be calculated from information extracted from metadata;

[0174] Surface reflectivity.

[0175] According to the surface reflectivity calculated above, the AOD value in the AOD inversion lookup table is obtained by reverse calculation.

[0176] S5 uses the spatiotemporal expansion strategy to expand the AODs of the first subset, the second subset, and the third subset respectively to obtain the expanded AODs, and obtains the AOD of the entire set of remote sensing data according to the expanded AODs and the AODs of the first subset, the second subset, and the third subset.

[0177] Specifically, step S5 includes:

[0178] S51 time-space expansion strategy includes time expansion strategy and space expansion strategy.

[0179] S52 performs time series expansion on the AODs of the first subset, the second subset, and the third subset respectively according to the time series expansion strategy to obtain a first time series simulation AOD, a second time series simulation AOD, and a third time series simulation AOD.

[0180] In one embodiment of the present invention, the timing expansion strategy is:

[0181] Step 1: Establish three time segment sequences according to the data acquisition time in the three classification sets. Each time segment is one month, and the sequence is arranged from recent to far.

[0182] Step 2: According to the data acquisition time of each data in the three subsets, a distribution diagram of the landing point of the subset in the corresponding time segment sequence is formed to obtain three distribution diagrams of the landing point;

[0183] Step 3, calculate the number of landing points in each time segment in a single landing point distribution diagram, and use the data corresponding to the time segments with the last 40% of the number of landing points as the time data set to be expanded for the corresponding subset;

[0184] Step 4: Taking the collection time of a single time data to be expanded of the time data set to be expanded as the intermediate node, selecting data within a week centered on the intermediate node in the corresponding classification set, and eliminating data with a quality score lower than 50%, to obtain a set of time candidate data sets;

[0185] Step 5: Repeat step 4 to obtain multiple sets of candidate time data sets for each time data set to be expanded;

[0186] Step 6: Calculate the AOD of all the time candidate data in a set of time candidate data sets, and take the average of all the AODs in the set of time candidate data sets and the AOD of the corresponding time data to be expanded as the candidate time series simulation AOD;

[0187] Step 7: Repeat step 6, and obtain multiple candidate time series simulation AODs for each subset. The multiple candidate time series simulation AODs are classified according to the affiliation of the subsets to obtain the first time series simulation AOD, the second time series simulation AOD, and the third time series simulation AOD.

[0188] Specifically, the timing expansion methods of the first subset, the second subset, and the third subset in the present invention are the same, and the timing expansion process of the first subset is taken as an example for description:

[0189] (1) The first classification set {C1,C2,...,C x} The data acquisition times of all the data are combined and uniformly divided into monthly units to establish a time segment sequence {PT1, PT2,..., PT t}. The beginning of the sequence is the most recent time segment, and they are arranged in descending order according to the adjacent time sequence.

[0190] (2) According to the data acquisition time of each data in the first subset {D1, D2,..., D y , y < x}, compare which time segment the data acquisition time of this data falls into. Start comparing from the first data D1. If the data acquisition time of D1 falls into the PT2 time segment, mark a landing point in the PT2 time segment. Compare and mark the subsequent data in this way until the comparison of the last data is completed, obtaining a landing point distribution diagram, which can be regarded as the time scatter diagram of the first subset. It reflects the data collection time law of the first subset, and it can be analyzed from this diagram which time segment has sufficient data in the first subset and which time segment lacks data.

[0191] (3) The landing point distribution diagram reflects the data collection time law of the first subset. It can be analyzed from this diagram which time segment has sufficient data in the first subset and which time segment lacks data, so as to find out the lacking part of the data for expansion. By calculating the number of landing points in each time segment of the landing point distribution diagram, arrange the time segments in descending order according to the number of landing points, and take the data corresponding to the last 40% of the time segments as the time data set to be expanded of the first subset {DT1, DT2,..., DT z , z < y}.

[0192] (4) Taking DT1 in the time data set to be expanded {DT1, DT2,..., DT z , z < y} as the intermediate node, with the data acquisition time of DT1 as the central node, find the data three days before and three days after this central node in the first classification set, and list the quality scores of these data. If the quality score is lower than 50%, then剔除 this data, and finally obtain a set of time candidate data sets {DT1 - 1} i , where i is a variable.

[0193] (5) Repeat the previous step to select time candidate data sets for the remaining {DT2,..., DT z} respectively. Then the time data set to be expanded will obtain a total of z sets of time candidate data sets {{DT1 - 1} i , {DT2 - 2} i ,..., {DT z - z} i}, where i is a variable, and the number of data in each set of time candidate data sets is different.

[0194] (6) In the z group of time candidate data sets {{DT1-1} i ,{DT2-2} i ,...,{DT z -z} i} select a set of temporal candidate datasets {DT1-1} i , calculate {DT1-1} i The AODs of all candidate time data in are obtained, and i AODs are obtained. The average of these i AODs and the AOD of DT1 is taken as the candidate time series simulation AOD of DT1.

[0195] (7) Repeat the previous step for the remaining {{DT2-2} i ,...,{DT z -z} i} also calculates the candidate timing simulation AOD, and finally obtains z candidate timing simulation AODs, and uses these z candidate timing simulation AODs as the first timing simulation AOD.

[0196] The present invention can expand the AOD of the three subsets from the time dimension through the above-mentioned time series expansion strategy, and the expanded time series simulation AOD is used as the AOD estimation of other data in the time dimension, and the time series simulation AOD has higher accuracy.

[0197] S53 spatially expands the AODs of the first subset, the second subset, and the third subset according to the spatial expansion strategy to obtain a first spatially simulated AOD, a second spatially simulated AOD, and a third spatially simulated AOD.

[0198] In one embodiment of the present invention, the space expansion strategy is:

[0199] The first step is to use the merged area of ​​the data acquisition range in each set of the first classification set, the second classification set, and the third classification set as their respective target areas, calculate the union of the data acquisition range of each subset, and use the mean of the AOD of all the data in the subset as the AOD of the corresponding union;

[0200] The second step is to compare the single union with the corresponding target area, and take the non-intersecting area as the area to be expanded of the corresponding subset;

[0201] Step 3: spatially interpolate the AOD of the single union to the corresponding area to be expanded with a step size of 3% until the area to be expanded of the corresponding subset is completely covered to obtain the AOD of the single area to be expanded;

[0202] Step 4: Repeat steps 2 and 3 to obtain AODs of three areas to be expanded, which are used as the first space simulation AOD, the second space simulation AOD, and the third space simulation AOD respectively.

[0203] The present invention expands the AOD from two dimensions, time and space, and can simulate the AOD of large-scale data with small-scale data to the greatest extent, taking into account the different characteristics of remote sensing data in time and space. The obtained simulated AOD can greatly represent the AOD of remote sensing data, thereby improving the accuracy of inversion.

[0204] S54 calculates the extended AOD in a weighted manner:

[0205] The AOD of the first subset, the first temporal simulation AOD, and the first spatial simulation AOD are calculated according to a first weighting method to obtain a first extended AOD.

[0206] The first weighting method is:

[0207]

[0208] stf1∈[1,F1],i1∈[1,n1],j1∈[1,m1],k1∈[1,h1],α1,β1,a,b∈[0,1]

[0209] in, is the f1th first extended AOD, F1 is the number of first extended AODs, α1 and β1 are weight coefficients, is the adjustment coefficient, n1 is the number of data in the first subset, m1 is the number of AODs simulated in the first time series, h1 is the number of AODs simulated in the first space, represents the AOD of the i1th data in the first subset, represents the j1th first time series simulation AOD, represents the k1th first spatial simulation AOD.

[0210] In this embodiment, a and b are used to adjust the deviation between the first time series simulation AOD and the AOD of the first subset in the time dimension, α1<β1, and when calculating the first extended AOD, the weight of the first spatial simulation AOD should be greater than the weight of the first time series simulation AOD.

[0211] The AOD of the second subset, the second temporal simulation AOD, and the second spatial simulation AOD are calculated according to a second weighting method to obtain a second extended AOD.

[0212] The second weighting method is:

[0213]

[0214] stf2∈[1,F2],i2∈[1,n2],j2∈[1,m2],k2∈[1,h2],α2,β2∈[0,1]

[0215] in, is the f2th second extended AOD, F2 is the number of second extended AODs, α2 and β2 are weight coefficients, n2 is the number of data in the second subset, m2 is the number of second temporal simulation AODs, h2 is the number of second spatial simulation AODs, represents the AOD of the i2th data in the second subset, represents the j2th second time series simulation AOD, represents the k2th second spatial simulation AOD.

[0216] In this embodiment, α1≈β1, and when calculating the second extended AOD, the weight of the second spatial simulation AOD should be not much different from the weight of the second temporal simulation AOD.

[0217] The AOD of the third subset, the third time series simulation AOD, and the third space simulation AOD are calculated according to a third weighting method to obtain a third extended AOD.

[0218] The third weighting method is:

[0219]

[0220] stf3∈[1,F3],i3∈[1,n3],j3∈[1,m3],k3∈[1,h3],α3,β3∈[0,1]

[0221] in, is the f3th third extended AOD, F3 is the number of third extended AODs, α3 and β3 are weight coefficients, n3 is the number of data in the third subset, m3 is the number of third time series simulation AODs, and h3 is the number of third space simulation AODs. represents the AOD of the i3th data in the third subset, represents the j3rd time series simulation AOD, represents the k3th third space simulation AOD.

[0222] In this embodiment, α1>β1, when calculating the third extended AOD, the weight of the third spatial simulation AOD should be smaller than the weight of the third temporal simulation AOD.

[0223] S55 combines the first extended AOD, the second extended AOD, the third extended AOD, the AOD of the first subset, the AOD of the second subset, and the AOD of the third subset as the AOD of the remote sensing data.

[0224] According to the characteristics of the scenes covered by the data in the three classification sets, the present invention proposes three weighting methods respectively. The first weighting method takes into account that most of the scenes in the first classification set are urban construction land, and the weight of the time dimension is not as heavy as that of the space dimension. Therefore, when calculating the first extended AOD, the first weighting formula is proposed taking into account this characteristic; the second weighting method takes into account that the scenes in the second classification set include both urban construction land and vegetation areas, so the weight of the time dimension and the weight of the space dimension should be almost the same; the third weighting method takes into account that most of the scenes in the third classification set are vegetation areas, and the vegetation will change significantly according to the seasons, so the weight of the time dimension should be greater than the weight of the space dimension.

[0225] The present invention designs three weighting methods and comprehensively considers the characteristics of different scenes. The calculated extended AOD can be more consistent with the original AOD of remote sensing data, improve the accuracy of simulation, and increase the accuracy of inversion. The subsequent atmospheric correction can achieve the effect of fine correction.

[0226] S6 uses the atmospheric radiation transfer model to establish an atmospheric correction lookup table, and performs a unified atmospheric correction on the remote sensing data according to the AOD based on the atmospheric correction lookup table to obtain the surface reflectivity of the remote sensing data.

[0227] Specifically, in this embodiment, the atmospheric correction lookup table parameters are set as follows:

[0228] 2 atmosphere types: mid-latitude summer and mid-latitude winter;

[0229] 1 aerosol type: continental aerosol;

[0230] 4 bands: 1-blue light band, 2-green light band, 3-red light band, 4-near infrared band;

[0231] 13 satellite zenith angles: 0-72, step size is 6, unit is degree;

[0232] 13 solar zenith angles: 0-72, step size is 6, unit is degree;

[0233] 19 relative azimuths: 0-180, step size is 10, unit is degree;

[0234] 14 AOD: values ​​are 0.0001, 0.1-1.0, 0.1-1.2, 0.1-1.5, 0.1-2.0, with a step size of 0.1;

[0235] 1 surface reflectivity: 0.1;

[0236] Output: atmospheric correction parameters q, w, e.

[0237] According to the AOD of remote sensing data and the atmospheric correction lookup table, the remote sensing data can be atmospherically corrected to obtain the surface reflectivity of the remote sensing data, thus completing the atmospheric correction process.

[0238] The above implementation modes are only used to illustrate the present invention, but not to limit the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the present invention. The patent protection scope of the present invention should be defined by the claims.

Claims

1. A remote sensing data atmospheric correction method, characterized in that: include: S1 obtains a set of remote sensing data and corresponding metadata, and preprocesses the remote sensing data; S2 calculates the normalized vegetation index of the remote sensing data, sets two thresholds, and classifies the remote sensing data according to the normalized vegetation index based on the two thresholds to obtain three classification sets, namely, the first classification set, the second classification set, and the third classification set; S3 screens the first classification set, the second classification set, and the third classification set respectively according to the metadata to obtain a first subset, a second subset, and a third subset; S4 calculates the AOD of the first subset, the second subset, and the third subset respectively according to the metadata and the atmospheric radiation transfer model; S5 uses a spatiotemporal expansion strategy to expand the AODs of the first subset, the second subset, and the third subset respectively to obtain an expanded AOD, and obtains the AOD of the entire set of remote sensing data according to the expanded AOD and the AODs of the first subset, the second subset, and the third subset; S6 uses the atmospheric radiation transfer model to establish an atmospheric correction lookup table, and performs a unified atmospheric correction on the remote sensing data according to the AOD based on the atmospheric correction lookup table to obtain the surface reflectivity of the remote sensing data.

2. The method according to claim 1, characterized in that Step S3 includes: S31 extracts the data acquisition range and data acquisition date from the metadata, and performs quality assessment on the remote sensing data according to the cloud detection results to obtain the quality score of the remote sensing data; S32 arranges the remote sensing data in chronological order from recent to far according to the data acquisition date, and divides them into sections with months as nodes to obtain multiple sets of section sets; S33 divides the remote sensing data according to the data acquisition range according to the preset divisions, and obtains multiple groups of range sets based on the coverage of the preset divisions reaching 60%; S34 sorts each group of segment sets and each group of range sets from high to low according to the quality scores, and cross-screens the sorted segment sets and the sorted range sets to obtain a screening set; S35 compares the screening set with the first classification set, the second classification set, and the third classification set: If the data in the screening set belongs to the first classification set, the data is classified into the first subset; If the data in the screening set belongs to the second classification set, the data is classified into the second subset; If the data in the screening set belongs to the third classification set, the data is classified into the third subset.

3. The method according to claim 2, characterized in that Step S34 includes: recording the sorted segment set as {{A i }} m , and record the sorted range set as {{B j }} n , where m is the number of segments, A i is the i-th segment data of the current group, n is the number of groups in the range set, B j is the j-th range data of the current group; Set m groups {A i The segment data with the top 50% quality scores in} are taken as the segment set; Set n groups {B j The range data with the top 30% quality scores in} are taken as the range set; The segment set and range set are merged and deduplicated to obtain the filter set.

4. The method according to claim 3, characterized in that Step S5 includes: S51 time-space expansion strategy includes time expansion strategy and space expansion strategy; S52 performs time series expansion on the AODs of the first subset, the second subset, and the third subset respectively according to the time series expansion strategy to obtain a first time series simulation AOD, a second time series simulation AOD, and a third time series simulation AOD; S53 spatially expands the AODs of the first subset, the second subset, and the third subset according to the spatial expansion strategy to obtain a first spatial simulation AOD, a second spatial simulation AOD, and a third spatial simulation AOD; S54 calculates the extended AOD in a weighted manner: Calculate the AOD of the first subset, the first temporal simulation AOD, and the first spatial simulation AOD according to a first weighting method to obtain a first extended AOD; The AOD of the second subset, the second temporal simulation AOD, and the second spatial simulation AOD are calculated according to a second weighting method to obtain a second extended AOD; The AOD of the third subset, the third time series simulation AOD, and the third space simulation AOD are calculated according to a third weighting method to obtain a third extended AOD; S55 combines the first extended AOD, the second extended AOD, the third extended AOD, the AOD of the first subset, the AOD of the second subset, and the AOD of the third subset as the AOD of the remote sensing data.

5. The method according to claim 4, characterized in that The timing expansion strategy is: Step 1: Establish three time segment sequences according to the data acquisition time in the three classification sets. Each time segment is one month, and the sequence is arranged from recent to far. Step 2: According to the data acquisition time of each data in the three subsets, a distribution diagram of the landing point of the subset in the corresponding time segment sequence is formed to obtain three distribution diagrams of the landing point; Step 3, calculate the number of landing points in each time segment in a single landing point distribution diagram, and use the data corresponding to the time segments with the last 40% of the number of landing points as the time data set to be expanded for the corresponding subset; Step 4: Taking the collection time of a single time data to be expanded of the time data set to be expanded as the intermediate node, selecting data within a week centered on the intermediate node in the corresponding classification set, and eliminating data with a quality score lower than 50%, to obtain a set of time candidate data sets; Step 5: Repeat step 4 to obtain multiple sets of candidate time data sets for each time data set to be expanded; Step 6: Calculate the AOD of all the time candidate data in a set of time candidate data sets, and take the average of all the AODs in the set of time candidate data sets and the AOD of the corresponding time data to be expanded as the candidate time series simulation AOD; Step 7: Repeat step 6, and obtain multiple candidate time series simulation AODs for each subset. The multiple candidate time series simulation AODs are classified according to the affiliation of the subsets to obtain the first time series simulation AOD, the second time series simulation AOD, and the third time series simulation AOD.

6. The method according to claim 4, characterized in that The space expansion strategy is: The first step is to use the merged area of ​​the data acquisition range in each set of the first classification set, the second classification set, and the third classification set as their respective target areas, calculate the union of the data acquisition range of each subset, and use the mean of the AOD of all the data in the subset as the AOD of the corresponding union; The second step is to compare the single union with the corresponding target area, and take the non-intersecting area as the area to be expanded of the corresponding subset; Step 3: spatially interpolate the AOD of the single union to the corresponding area to be expanded with a step size of 3% until the area to be expanded of the corresponding subset is completely covered to obtain the AOD of the single area to be expanded; Step 4: Repeat steps 2 and 3 to obtain AODs of three areas to be expanded, which are used as the first space simulation AOD, the second space simulation AOD, and the third space simulation AOD respectively.

7. The method according to claim 4, characterized in that: The first weighting method is: stf1∈[1,F1],i1∈[1,n1],j1∈[1,m1],k1∈[1,h1],α1,β1,a,b∈[0,1] in, is the f1th first extended AOD, F1 is the number of first extended AODs, α1 and β1 are weight coefficients, a and b are adjustment coefficients, n1 is the number of data in the first subset, m1 is the number of first time series simulation AODs, h1 is the number of first space simulation AODs, represents the AOD of the i1th data in the first subset, represents the j1th first time series simulation AOD, represents the k1th first spatial simulation AOD; The second weighting method is: stf2∈[1,F2],i2∈[1,n2],j2∈[1,m2],k2∈[1,h2],α2,β2∈[0,1] in, is the f2th second extended AOD, F2 is the number of second extended AODs, α2 and β2 are weight coefficients, n2 is the number of data in the second subset, m2 is the number of second time series simulation AODs, h2 is the number of second space simulation AODs, represents the AOD of the i2th data in the second subset, represents the j2th second time series simulation AOD, represents the k2th second space simulation AOD; The third weighting method is: stf3∈[1,F3],i3∈[1,n3],j3∈[1,m3],k3∈[1,h3],α3,β3∈[0,1] in, is the f3th third extended AOD, F3 is the number of third extended AODs, α3 and β3 are weight coefficients, n3 is the number of data in the third subset, m3 is the number of third time series simulation AODs, and h3 is the number of third space simulation AODs. represents the AOD of the i3th data in the third subset, represents the j3rd time series simulation AOD, represents the k3th third space simulation AOD.

8. The method according to claim 1, characterized in that: Step S1 includes: S11 obtains the band value of each pixel of each remote sensing data by calculation; S12 obtains historical remote sensing data, uses the long-term historical remote sensing data to construct a clear sky background field containing a blue light band, and calculates a cloud detection threshold based on the clear sky background field containing a blue light band and the blue light band reflectance of each pixel of the remote sensing data; S13 uses the cloud detection threshold to perform cloud detection on the remote sensing data, removes cloud and cloud shadow pixels, and obtains pre-processed remote sensing data.

9. The method according to claim 8, characterized in that Step S2 includes: S21 calculates the normalized vegetation index of remote sensing data. The calculation formula is as follows: Among them, NDVI is the normalized vegetation index, the NDVI value range is [-1,1], ρ NIR is the near infrared band value, ρ RED is the reflectivity of red light band; S22 sets a first threshold value ρ1 and a second threshold value ρ2, the first threshold value ρ1 is less than the second threshold value ρ2, and the first threshold value ρ1 and the second threshold value ρ2 divide the value range of NDVI [-1,1] into three value segments, namely, a first value segment [-1,ρ1), a second value segment [ρ1,ρ2], and a third value segment [ρ2,1]; S23 compares the NDVI value of remote sensing data with three value ranges: If the NDVI value of a remote sensing data falls within the range of the first value segment, the remote sensing data is classified into the first classification set; If the NDVI value of a remote sensing data falls within the range of the second value segment, the remote sensing data is classified into the second classification set; If the NDVI value of a remote sensing data falls within the range of the third value segment, the remote sensing data is classified into the third classification set.

10. The method according to claim 1, characterized in that Step S4 includes: S41 extracts calibration information and solar geometry information from metadata; S42 calculates the apparent reflectance of each data in the first subset, the second subset, and the third subset respectively according to the calibration information and the solar geometry information; S43 uses the atmospheric radiation transfer model to establish an AOD inversion lookup table; S44 calculates the AOD of the first subset, the second subset, and the third subset respectively according to the metadata, the AOD inversion lookup table, and the apparent reflectivity.