Hyperspectral remote sensing data atmospheric correction method and system based on spectral smoothness

Abstract

The invention provides a hyperspectral remote sensing data atmospheric correction method based on spectral smoothness. The hyperspectral remote sensing data atmospheric correction method comprises the steps of 1, generating a representative atmospheric parameter lookup table and atmospheric roof apparent spectral radiance; 2, constructing a deep learning network model to estimate the atmospheric aerosol optical thickness and the atmospheric column water vapor content initial value; and step 3, estimating the atmospheric aerosol optical thickness and the atmospheric column water vapor content based on the reflectivity spectrum smoothness, and performing atmospheric correction. According to the method provided by the invention, a novel method for estimating the optical thickness of the pixel-level atmospheric aerosol and the moisture content of the atmospheric column by using the hyperspectral remote sensing image is developed, and a novel method and system for obtaining the surface spectral reflectivity of the pixel by performing atmospheric correction are developed; therefore, a key method is provided for atmospheric correction of spaceborne hyperspectral remote sensing data in the wavelength range of 0.4-2.5 microns.

Description

Technical Field

[0001] The present invention relates to a technical method and system for atmospheric correction of hyperspectral remote sensing data using the smoothness of the reflectivity spectrum between 0.4-2.5 μm wavelengths, and belongs to the field of quantitative remote sensing parameter inversion. The atmospheric correction technology of hyperspectral remote sensing data inverted by the present invention has good applicability and high accuracy, and can be applied to atmospheric correction of hyperspectral remote sensing between 0.4-2.5 μm wavelengths to obtain surface reflectivity. Background Art

[0002] When electromagnetic wave signals in the wavelength range of 0.4-2.5μm penetrate the atmosphere, they will be attenuated due to absorption and scattering by the atmosphere. The radiance observed by the remote sensor cannot fully reflect the surface reflection ability. Therefore, it is often necessary to input information such as atmospheric aerosol optical thickness and water vapor content to calculate the radiance observed by the remote sensor to the actual reflectivity of the ground. This process is called atmospheric correction of remote sensing data in this wavelength range.

[0003] Aerosol Optical Depth (AOD) and Precipitable Water Vapor (PWV) are the key parameters of common atmospheric correction. For multispectral remote sensing data with a wavelength of 0.4-2.5 μm, the dark target method is often used to extract targets such as dense vegetation and clean water bodies, assuming that there are dark target pixels in the remote sensing image, or to invert with the help of polarization imaging; the current conventional means of monitoring atmospheric water vapor content include remote sensing inversion of meteorological remote sensing satellite and radar data, radiosonde monitoring, ground-based water vapor radiometer or GPS monitoring, and estimation of direct solar radiation observed by sun photometers. Among them, the most widely used method for obtaining atmospheric water vapor content is to estimate atmospheric water vapor content using the water vapor absorption channel at 940 nm. However, hyperspectral data includes rich surface and atmospheric radiation information, and its spectral information can be used to establish a more convenient integrated method for atmospheric parameter extraction and atmospheric correction. Summary of the invention

[0004] The technical problem to be solved by the present invention is: how to directly use the hyperspectral remote sensing data in the wavelength range of 0.4-2.5μm to estimate the atmospheric parameters and realize the atmospheric correction of the data to obtain the real ground reflectivity.

[0005] At present, the atmospheric correction of optical hyperspectral remote sensing data first needs to estimate the atmospheric aerosol optical thickness using dark target pixels such as dense vegetation or clear water bodies. However, this method will fail for images without dark targets. At the same time, the AOD of a limited number of pixels is used to replace the value of the entire image, which cannot reflect the spatial difference of AOD, resulting in errors in its subsequent atmospheric correction process. In view of the shortcomings of the prior art, the present invention discloses a method for atmospheric correction of hyperspectral remote sensing data with spectral smoothness, which directly estimates the AOD and PWV of each pixel, and implements atmospheric correction of the pixel to obtain the spectral reflectance, thereby getting rid of the dependence on dark targets and the influence of spatial heterogeneity of atmospheric parameters.

[0006] The innovation of the present invention lies in: (1) Proposing to use a deep learning model to estimate the initial values ​​of aerosol optical thickness and atmospheric column water vapor content from 0.4-2.5μm hyperspectral radiance data. (2) Based on the initial values ​​of AOD and PWV, an index value range matrix of the two is established, and each group of AOD and PWV is used to perform atmospheric correction of hyperspectral pixels to obtain spectral reflectance, and the smoothness of the spectral reflectance is calculated. (3) A two-dimensional Gaussian surface function is constructed, and the AOD and PWV corresponding to the minimum value of the smoothness value are the true atmospheric parameters of the pixel, and the corresponding spectral reflectance is the reflectance after atmospheric correction.

[0007] The technical solution adopted by the present invention is as follows:

[0008] A method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness comprises the following steps:

[0009] Step 1: Generate representative atmospheric parameter lookup table and apparent spectral radiance of top of atmosphere. Specifically include:

[0010] Based on the atmospheric radiation transmission software, the spectral response function of the hyperspectral sensor with different atmospheric profile types, aerosol optical thickness, atmospheric column water vapor content, solar incidence angle and wavelength range of 0.4-2.5μm is simulated, and the atmospheric parameter lookup table at different bands under the above conditions is established.

[0011] (1) Atmospheric profile types: 5 categories: tropical atmospheric profile, mid-latitude summer atmospheric profile, mid-latitude winter atmospheric profile, subpolar summer atmospheric profile and subpolar winter atmospheric profile.

[0012] (2) Aerosol optical thickness value, there are 21 levels: 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35, 0.40, 0.45, 0.50, 0.55, 0.60, 0.65, 0.70, 0.75, 0.80, 0.85, 0.90, 0.95, 1.0, 2.0.

[0013] (3) Water vapor content in the atmospheric column, 10 levels: 0.1×WV max , 0.2×WV max , 0.3×WV max , 0.4×WV max , 0.5×WV max , 0.6×WV max , 0.7×WV max , 0.8×WV max , 0.9×WV max , 1.0×WV max Among them, WV max Indicates the maximum water vapor content of the atmospheric profile.

[0014] (4) Solar incidence angle, there are 8 levels: 0°, 10°, 20°, 30°, 40°, 50°, 60° and 70°.

[0015] (5) The spectral response function of a certain band of the hyperspectral spectrum is expressed by a Gaussian function:

[0016]

[0017] Among them, λ c and FWHM are the central wavelength and band width of a certain band, respectively. i ) is the spectral response rate of the simulated wavelength, and the value is between 0 and 1.0.

[0018] The lookup table and reflectance data of different surface types are used to simulate the apparent radiance of the spectrum of different combinations of hyperspectral sensors at the top of the atmosphere based on the atmospheric radiation transfer equation. The surface types should include vegetation, bare soil, rock, water, ice and snow, etc., and the number of various types should be equal, with a total of no less than 20 types.

[0019] The atmospheric radiation transfer equation is expressed as:

[0020]

[0021] Among them, L TOA is the top-of-atmosphere radiance, L path is the atmospheric radiation, F is the parameter related to solar energy and atmosphere, s is the atmospheric downward hemispheric albedo, ρ is the surface reflectivity, and each parameter changes with the band. Among them, the atmospheric parameters required for the atmospheric correction involved in the present invention are L path , F and s.

[0022] Step 2: Construct a deep learning network model to estimate the initial values ​​of atmospheric aerosol optical thickness and atmospheric column water vapor content. Specifically include:

[0023] Based on the simulated aerosol optical thickness, atmospheric column water vapor content and the spectral apparent radiance data, a deep learning network model (DNN) consisting of a long short-term memory network (LSTM) and a convolutional neural network (CNN) was established to estimate the atmospheric aerosol optical thickness and atmospheric column water vapor content from the spectral radiance data. The input of the model is the spectral radiance of a single pixel, and the output is the atmospheric aerosol optical thickness and atmospheric column water vapor content corresponding to the pixel.

[0024] By using the observation longitude and latitude, observation time and observed spectral apparent radiance of hyperspectral remote sensing data and inputting the established deep learning network model, the initial values ​​of aerosol optical thickness and atmospheric column water vapor content (AOD0 and PWV0) of each pixel can be obtained.

[0025] Step 3: Estimation of atmospheric aerosol optical depth and atmospheric column water vapor content based on reflectance spectrum smoothness and atmospheric correction, including:

[0026] Based on the initial values ​​of aerosol optical thickness and atmospheric column water vapor content, the search value ranges of the two parameters are set, namely ΔAOD and ΔPWV, and then the index value range matrix of AOD and PWV is formed with ΔAOD / 10 as the interval between AOD0-ΔAOD and AOD0+ΔAOD, and with ΔPWV / 10 as the interval between PWV0-ΔPWV and PWV0+ΔPWV; based on the lookup table, the local lookup table of atmospheric parameters corresponding to each pair of elements (AODx, PWVx) in the matrix is ​​established by interpolation. Then the spectral reflectance smoothness is calculated, and the minimum value of the smoothness is used to determine the final aerosol optical thickness and atmospheric water vapor content. Specifically including:

[0027] Combined with the (AOD x , PWV x ) matrix, the corresponding surface spectral reflectance is calculated from the spectral apparent radiance observed by the pixel based on the atmospheric radiation transfer equation (Formula (2)), and then the corresponding curve smoothness S is calculated. x , PWV x ) and find the smoothness value S min0 The minimum value of AOD corresponds to min0 and PWV min0 .

[0028] Surface reflectivity calculation formula:

[0029]

[0030] Smoothness calculation method:

[0031]

[0032] Among them, ρ i , i-1 and ρ i+1 The surface reflectances of the i-th, i-1-th and i+1-th bands of the hyperspectral sensor, respectively, where i increases one by one from 2 to the number of hyperspectral bands - 1.

[0033] Then, in the XYZ three-dimensional coordinate system composed of AOD, PWV and smoothness, the point (AOD min0 , PWV min0 , S min0 ) as the center, a two-dimensional Gaussian surface function with AOD and PWV as independent variables and smoothness as dependent variable is fitted in a smaller range, and the global minimum value S of smoothness is obtained by derivation of the function. min Then the corresponding AOD and PWV are used as the actual atmospheric aerosol optical thickness AOD of the pixel. est and atmospheric column water vapor content PWV est .

[0034] Two-dimensional Gaussian surface function of smoothness:

[0035]

[0036] Where G is the amplitude of the Gaussian distribution, σ x and σ y Respectively (AOD x , PWV x ) matrix.

[0037] Then, based on the local lookup table of atmospheric parameters, the atmospheric parameters corresponding to (AODest, PWVest) are obtained by interpolation, and the surface spectral reflectance (Formula (3)) is calculated from the apparent spectral radiance of the pixel using the atmospheric radiation transfer equation (Formula (2)). Finally, the construction of an integrated method for atmospheric parameter extraction and atmospheric correction is realized.

[0038] The present invention proposes a hyperspectral remote sensing data atmospheric correction system based on spectral smoothness, which comprises:

[0039] The estimation model building module is used to simulate the atmospheric parameter lookup table and the spectral radiance of the top of the atmosphere under different atmospheric conditions and different angles based on the atmospheric radiation transfer software, and to establish a deep learning network model that estimates the atmospheric aerosol optical thickness and atmospheric column water vapor content from the spectral radiance data;

[0040] The data processing module is used to read the actual hyperspectral remote sensing data and extract the spectral radiance data of each pixel;

[0041] The atmospheric correction module is used to estimate the atmospheric aerosol optical thickness and the initial value of atmospheric column water vapor content from the spectral radiance data, and to determine the final aerosol optical thickness and atmospheric column water vapor content of the pixel using the smoothness method and to obtain the pixel surface spectral reflectance.

[0042] Among them, the "estimation model establishment module" is mainly used to build a deep learning network based on the characteristics of hyperspectral sensors; the "data processing module" inputs the processed apparent radiance data of each pixel into the "atmospheric correction module", and the "atmospheric correction module" calls the deep learning network constructed by the "estimation model establishment module" to obtain the initial values ​​of atmospheric aerosol optical thickness and atmospheric column water vapor content, and further uses the smoothness method to determine the final aerosol optical thickness and atmospheric column water vapor content of the pixel and obtain the pixel surface spectral reflectance.

[0043] The beneficial effects of the present invention are as follows:

[0044] Based on deep learning methods and smoothness methods, the present invention develops a new method for estimating pixel-level atmospheric aerosol optical thickness and atmospheric column water vapor content using hyperspectral remote sensing images, and implements atmospheric correction to obtain a new method and system for pixel surface spectral reflectance, thereby providing a key method for atmospheric correction of satellite-borne hyperspectral remote sensing data in the 0.4-2.5μm wavelength range. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Flow chart of the method of the present invention.

[0046] Figure 2 is the central wavelength of each band and the atmospheric transmittance.

[0047] Figure 3 Look up a diagram of the structure for atmospheric parameters.

[0048] Figure 4a , Figure 4b , Figure 4c This is an example of the spectral atmospheric parameters Lpath, F and S corresponding to WV=1g / cm2 and AOD=0.2 in the lookup table.

[0049] Figure 5 It is a schematic diagram of the spectral reflectance spectrum of vegetation, soil and other ground objects.

[0050] Figure 6 Schematic diagram of the apparent spectral radiance at the top of the atmosphere.

[0051] Figure 7a is the AOD difference histogram.

[0052] Figure 7b This is a graph showing how the AOD error changes with the AOD size.

[0053] Figure 8a is the WV difference histogram.

[0054] Figure 8b This is a graph showing the variation of WV error with WV size. DETAILED DESCRIPTION

[0055] The present invention designs a method and system for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness. In order to more clearly illustrate the purpose, technical solutions and advantages of the present invention, the present invention will be further elaborated in detail below in conjunction with the accompanying drawings. The described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limitations on the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0056] Step 1: Sensor Information and Characteristics

[0057] In order to test the method, according to the existing instrument description, in the visible-near infrared (VNIR, 0.4-1.0μm) range, the central wavelength starts from 0.4μm, and a total of 121 bands with a spectral resolution (two-by-two intervals) of 5nm and a band width FWHM of 5nm are designed; in the short-wave infrared (1.0-2.5μm) range, the central wavelength starts from 1.0μm, and a total of 150 bands with a spectral resolution of 10nm and a band width FWHM of 10nm are designed. Then there are a total of 271 bands in the range of 0.4-2.5μm. Figure 2 The central wavelength position of each band and its corresponding atmospheric transmittance under the standard mid-latitude atmospheric profile are displayed. It can be seen that there are multiple atmospheric non-window channels with atmospheric transmittance close to 0. Then these channel data will not be used in the subsequent spectral smoothness calculation.

[0058] Step 2: Establishment of hyperspectral atmospheric parameter lookup table

[0059] Based on the atmospheric profile type, aerosol optical thickness, atmospheric column water vapor content, and solar incidence angle conditions set above, the atmospheric radiation transfer model was used to simulate the three atmospheric parameters Lpath, F, and S in 271 bands under different atmospheric conditions, and a parameter lookup table was formed. Figure 3 The basic format of the lookup table is given in the figure, where the first line is the water vapor content in g / cm2, the second line is the aerosol optical depth AOD, and each line is the value of Lpath, F and S in each band under various combinations of water vapor content and AOD.

[0060] Figure 4a , Figure 4b , Figure 4c Given the standard atmospheric profile in mid-latitude summer, the solar zenith angle is 30°, and WV = 1g / cm 2 , AOD=0.2, look up the atmospheric parameters Lpath, F and S instances of each band in the table. Figure 4a , Figure 4b , Figure 4c Lookup table WV = 1g / cm 2 , AOD = 0.2 corresponding to the spectral atmospheric parameters Lpath, F and S examples; Based on the constructed atmospheric parameter lookup table, combined with the spectral reflectance curve of typical objects (such as Figure 5 As shown), combined with the atmospheric radiation transfer equation The corresponding apparent radiance of the top of the atmosphere can be simulated (such as Figure 6 In this example, the apparent spectral radiance of the top of the atmosphere under 5808 conditions is simulated.

[0061] Step 3: Atmospheric parameter estimation and atmospheric correction

[0062] Based on the formed apparent spectral radiance dataset, the atmospheric aerosol optical thickness and water vapor content are estimated using the above-mentioned deep learning network and spectral smoothing method.

[0063] Figure 7a and Figure 7b The histogram of the difference between the true value and the inverted value of aerosol optical thickness (Delta AOD) and the variation of the inverted aerosol optical thickness error (AOD RMSE) with the value of AOD itself are given respectively. It can be seen that the AOD error is mainly concentrated around 0.0; the inversion error under different AOD values ​​is between 0.04-0.073, and the overall error is 0.048. The minimum and maximum values ​​of the error are reached at AOD = 0.5 and 2.0 respectively. Overall, the AOD inversion value is close to the true value and has a high accuracy.

[0064] Similarly, Figure 8a and 8b The histogram of the difference between the true value and the inverted value of atmospheric water vapor content (Delta WV) and the change of the inverted atmospheric water vapor content error (WVRMSE) with the value of WV itself are given. It can be seen that the WV error is mainly concentrated around 0.0; the inversion error under different WV values ​​is between 0.10-0.52g / cm 2 The overall error is 0.24g / cm 2 . At WV = 1.5 g / cm 2 and 6g / cm 2 The minimum and maximum values ​​of the error are reached at . Overall, the WV inversion value is closer to the true value and has a higher accuracy.

[0065] The review results show that the hyperspectral remote sensing data atmospheric parameter acquisition method based on deep learning and spectral smoothness proposed in the present invention can obtain high-precision atmospheric aerosol optical thickness and water vapor content and subsequent surface reflectivity.

Claims

1. A method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness, characterized in that: The following steps are involved: Step 1: Generate representative atmospheric parameter lookup table and top of atmosphere apparent spectral radiance; Step 2: Construct a deep learning network model to estimate the initial values ​​of atmospheric aerosol optical thickness and atmospheric column water vapor content; Step 3: Estimation of atmospheric aerosol optical depth and atmospheric column water vapor content based on reflectance spectrum smoothness and atmospheric correction.

2. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 1, characterized in that: In step one, the spectral response function of the hyperspectral sensor with different atmospheric profile types, aerosol optical thickness, atmospheric column water vapor content, solar incidence angle and wavelength range of 0.4-2.5μm is simulated based on the atmospheric radiation transfer software, and a lookup table of atmospheric parameters in different bands under the above conditions is established.

3. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 2, characterized in that: Specifically: (1) Atmospheric profile types, which include five categories: tropical atmospheric profile, mid-latitude summer atmospheric profile, mid-latitude winter atmospheric profile, subpolar summer atmospheric profile, and subpolar winter atmospheric profile; (2) Aerosol optical thickness value, a total of 21 levels: 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35, 0.40, 0.45, 0.50, 0.55, 0.60, 0.65, 0.70, 0.75, 0.80, 0.85, 0.90, 0.95, 1.0, 2.0; (3) Water vapor content in the atmospheric column, 10 levels: 0.1×WV max , 0.2×WV max , 0.3×WV max , 0.4×WV max , 0.5×WV max , 0.6×WV max , 0.7×WV max , 0.8×WV max , 0.9×WV max , 1.0×WV max Among them, WV max Indicates the maximum water vapor content of the atmospheric profile; (4) Solar incidence angle, 8 levels in total: 0°, 10°, 20°, 30°, 40°, 50°, 60° and 70°; (5) The spectral response function of a certain band of the hyperspectral spectrum is expressed by a Gaussian function: Among them, λ c and FWHM are the central wavelength and band width of a certain band, respectively. i ) is the spectral response rate of the simulated wavelength, and the value is between 0 and 1.

0.

4. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 3, characterized in that: Using the lookup table and reflectance data of different surface types, simulate the apparent radiance of the spectrum of different combinations of hyperspectral sensors at the top of the atmosphere based on the atmospheric radiation transfer equation. The surface types should include vegetation, bare soil, rock, water, ice and snow, and the number of various types should be equal, with no less than 20 types overall. The atmospheric radiation transfer equation is expressed as: Among them, L TOA is the top-of-atmosphere radiance, L path is the atmospheric radiation, F is the parameter related to solar energy and atmosphere, s is the atmospheric downward hemispheric albedo, ρ is the surface reflectivity, and each parameter changes with the band.

5. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 1, characterized in that: In step 2, based on the aerosol optical thickness, atmospheric column water vapor content and spectral apparent radiance data used in the simulation, a deep learning network model DNN consisting of a long short-term memory network LSTM and a convolutional neural network CNN is established to estimate the atmospheric aerosol optical thickness and atmospheric column water vapor content from the spectral radiance data; the input of the model is the spectral radiance of a single pixel, and the output is the atmospheric aerosol optical thickness and atmospheric column water vapor content corresponding to the pixel.

6. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 1, characterized in that: The observation longitude and latitude, observation time and observed spectral apparent radiance of hyperspectral remote sensing data are used as input into the deep learning network model to obtain the initial values ​​of aerosol optical thickness and atmospheric column water vapor content AOD0 and PWV0 for each pixel.

7. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 1, characterized in that: In step three, based on the initial values ​​of aerosol optical thickness and atmospheric column water vapor content, the retrieval value ranges of the two parameters are set, namely ΔAOD and ΔPWV, and then the index value range matrix of AOD and PWV is formed with ΔAOD / 10 as the interval between AOD0-ΔAOD and AOD0+ΔAOD, and with ΔPWV / 10 as the interval between PWV0-ΔPWV and PWV0+ΔPWV; based on the lookup table, a local lookup table of atmospheric parameters corresponding to each pair of elements (AODx, PWVx) in the matrix is ​​established by interpolation; then the spectral reflectance smoothness is calculated, and the minimum value of the smoothness is used to determine the final aerosol optical thickness and atmospheric water vapor content.

8. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 7, characterized in that: Combined with the (AOD x , PWV x ) matrix, calculate the corresponding surface spectral reflectance from the spectral apparent radiance observed by the pixel based on the atmospheric radiation transfer equation, and then calculate the corresponding curve smoothness S; In the acquisition matrix, each (AOD x , PWV x ) and find the smoothness value S min0 The minimum value of AOD corresponds to min0 and PWV min0 ; Surface reflectivity calculation formula: Smoothness calculation method: Among them, ρ i , i-1 and ρ i+1 The surface reflectances of the i-th, i-1-th and i+1-th bands of the hyperspectral sensor, respectively, where i increases one by one from 2 to the number of hyperspectral bands - 1.

9. The method for atmospheric correction of hyperspectral remote sensing data based on spectral smoothness according to claim 8, characterized in that: In the XYZ three-dimensional coordinate system composed of AOD, PWV and smoothness, the point (AOD min0 , PWV min0 , S min0 ) as the center, a two-dimensional Gaussian surface function with AOD and PWV as independent variables and smoothness as the dependent variable is fitted within the range of AOD and PWV, and the global minimum value S of smoothness is obtained by derivation of the function. min Then the corresponding AOD and PWV are used as the actual atmospheric aerosol optical thickness (AOD) of the pixel. est and atmospheric column water vapor content PWV est ; Two-dimensional Gaussian surface function of smoothness: Where G is the amplitude of the Gaussian distribution, σ x and σ y Respectively (AOD x , PWV x ) matrix.

10. A hyperspectral remote sensing data atmospheric correction system based on spectral smoothness, characterized in that: include: The estimation model building module is used to simulate the atmospheric parameter lookup table and the spectral radiance of the top of the atmosphere under different atmospheric conditions and different angles based on the atmospheric radiation transfer software, and to establish a deep learning network model that estimates the atmospheric aerosol optical thickness and atmospheric column water vapor content from the spectral radiance data; The data processing module is used to read the actual hyperspectral remote sensing data and extract the spectral radiance data of each pixel; The atmospheric correction module is used to estimate the atmospheric aerosol optical thickness and the initial value of atmospheric column water vapor content from the spectral radiance data, and to determine the final aerosol optical thickness and atmospheric column water vapor content of the pixel using the smoothness method and to obtain the pixel surface spectral reflectance.

Images