A General Rayleigh Scattering Calculation Method and System for Water Color Remote Sensing

Abstract

The present invention provides a general Rayleigh scattering calculation method and system for water color remote sensing. First, a Rayleigh scattering data set is formed based on observation conditions and atmospheric parameter data, and then this data set is input into a machine learning method for training to obtain a Rayleigh scattering prediction model. Finally, the observation conditions and atmospheric parameter data of each band and each pixel of the target remote sensing image are obtained and input into the trained prediction model to calculate the Rayleigh scattering of each pixel and each band of the target remote sensing image. The method and system disclosed by the present invention are general and not specific to a particular sensor, and are applicable to all existing and future water color remote sensing sensors. At the same time, they have the advantages of high efficiency and high precision, and avoid the errors brought by the lookup table method in the linear interpolation process.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing, and particularly relates to a general Rayleigh scattering calculation method and system for water color remote sensing. Background Art

[0002] The calculation of Rayleigh scattering radiance is a crucial link in the atmospheric correction of remote sensing images, and it is usually achieved by using the single-scattering algorithm or solving the radiative transfer equation. Internationally, for the radiative transfer problems of the atmosphere and the ocean, a variety of numerical calculation methods have been developed, including the discrete ordinate method, the invariant imbedding method, the finite element method, the successive order method, and the Monte Carlo simulation method, etc. Based on the radiative transfer theory and numerical calculation methods, radiative transfer models such as DISORT, 6S, PolRadTran, MOMO, COART, and PCOART have been developed at home and abroad.

[0003] In water color remote sensing, since the water body signal is weak and extremely vulnerable to the influence of the atmosphere, in order to extract high-precision water-leaving radiation information from remote sensing images, the error in the calculation of Rayleigh scattering radiance must be controlled within 1%. Therefore, the calculation of Rayleigh scattering radiance needs to be achieved by solving an accurate radiative transfer equation.

[0004] However, the solution process of the radiative transfer equation is complex and time-consuming. In the actual atmospheric correction of water color satellite remote sensing data, in order to improve the operation efficiency, the correction of Rayleigh scattering usually adopts a look-up table method based on radiative transfer simulation, that is, for a specific water color satellite sensor, the radiative transfer equation under different solar angles and observation angles and other conditions in each band is solved in advance, and a look-up table file is generated according to the calculation results. In image processing, according to the solar and observation geometric conditions of each pixel at the transit time, the most similar pre-calculated result is found in the look-up table, and then the Rayleigh scattering radiance value corresponding to each pixel in each band is obtained through linear interpolation. However, this method has certain limitations. On the one hand, the look-up table is designed for a specific remote sensor and does not have universality; on the other hand, the use of the look-up table cannot avoid the error introduced by linear interpolation, which affects the calculation accuracy. Therefore, there is currently a lack of an efficient, fast, and general Rayleigh scattering contribution calculation method applicable to various water color satellite sensors. Summary of the Invention

[0005] The present invention provides a general Rayleigh scattering calculation method and system for water color remote sensing to solve the problem that the existing look-up table method is not universal.

[0006] In order to solve the above technical problems, the embodiments of the present invention disclose the following technical solutions:

[0007] One aspect of the present invention provides a general Rayleigh scattering calculation method for water color remote sensing, including:

[0008] Input multiple sets of preset observation conditions and atmospheric parameter data into the radiative transfer model PCOART respectively, and calculate the corresponding Rayleigh scattering Stokes vector data. The observation condition data includes at least the solar zenith angle, relative azimuth, and observation zenith angle, and the atmospheric parameter data includes at least the Rayleigh scattering optical depth of atmospheric molecules.

[0009] Train an XGBoost machine learning model based on the observation conditions, atmospheric parameter data, and the corresponding Rayleigh scattering Stokes vector data to obtain a Rayleigh scattering prediction model.

[0010] Obtain the observation conditions and atmospheric parameter data of the target remote sensing image and input them into the Rayleigh scattering prediction model to obtain the Rayleigh scattering Stokes vector data of each pixel in the target remote sensing image.

[0011] Optionally, the step of inputting multiple sets of preset observation conditions and atmospheric parameter data into the radiative transfer model PCOART respectively and calculating the corresponding Rayleigh scattering Stokes vector data includes:

[0012] Take equal-step values for the Rayleigh scattering optical depth values of atmospheric molecules at a first preset step within a first set range to obtain a set of Rayleigh scattering optical depth values of atmospheric molecules.

[0013] Take equal-step values for the solar zenith angle at a second preset step within a second set range to obtain a set of solar zenith angle values.

[0014] Take equal-step values for the relative azimuth at a third preset step within a third set range to obtain a set of relative azimuth values.

[0015] Pre-set the value of the observation zenith angle.

[0016] Based on the set of Rayleigh scattering optical depth values of atmospheric molecules, the set of solar zenith angle values, the set of relative azimuth values, and the preset observation zenith angle, combine different observation conditions and atmospheric parameters to obtain multiple sets of different observation conditions and atmospheric parameter data.

[0017] Input each set of observation conditions and atmospheric parameter data into the radiative transfer model PCOART respectively to obtain the corresponding Rayleigh scattering Stokes vector, which is composed of an I component, a Q component, and a U component.

[0018] Optionally, the step of combining different observation conditions and atmospheric parameters based on the set of Rayleigh scattering optical depth values of atmospheric molecules, the set of solar zenith angle values, the set of relative azimuth values, and the preset observation zenith angle to obtain multiple sets of different observation conditions and atmospheric parameter data includes:

[0019] Arrange and combine the values in the set of Rayleigh scattering optical thickness values of atmospheric molecules, the set of solar zenith angle values, and the set of relative azimuth angle values with the value of the observation zenith angle to form multiple sets of observation conditions and atmospheric parameter data.

[0020] Optionally, training an XGBoost machine learning model based on the observation conditions, atmospheric parameter data, and the corresponding Rayleigh scattering Stokes vector data to obtain a Rayleigh scattering prediction model includes:

[0021] Divide all sets of observation conditions, atmospheric parameter data, and the corresponding Rayleigh scattering Stokes vector data into a training set and a test set according to a preset ratio;

[0022] Use the Rayleigh scattering optical thickness of atmospheric molecules, the solar zenith angle, the relative azimuth angle, and the observation zenith angle as input features, and use the I component, Q component, and U component of the Rayleigh scattering Stokes vector as output features, and train the XGBoost machine learning model with the training set to obtain a Rayleigh scattering prediction model;

[0023] Use the test set to test the trained Rayleigh scattering prediction model.

[0024] Optionally, obtaining the observation conditions and atmospheric parameter data of the target remote sensing image and inputting them into the Rayleigh scattering prediction model to obtain the Rayleigh scattering Stokes vector data of each pixel in the target remote sensing image includes:

[0025] Obtain the Rayleigh scattering optical thickness values of atmospheric molecules in each band of the target remote sensing image;

[0026] Obtain the solar zenith angle, relative azimuth angle, and observation zenith angle of each pixel in the target remote sensing image;

[0027] Input the Rayleigh scattering optical thickness values of atmospheric molecules in each band and the solar zenith angle, relative azimuth angle, and observation zenith angle of each pixel into the Rayleigh scattering prediction model to obtain the Rayleigh scattering Stokes vector data corresponding to each pixel in each band of the target remote sensing image.

[0028] Optionally, obtaining the Rayleigh scattering optical thickness values of atmospheric molecules in each band of the target remote sensing image includes:

[0029] Calculate the equivalent Rayleigh scattering optical thickness value of each band using the following formula:

[0030]

[0031] where band is the band of the remote sensing sensor, is the wavelength of the band, is the Rayleigh scattering optical thickness value of atmospheric molecules in the band, is the solar irradiance at the average Earth - Sun distance obtained in advance, which is obtained from the remote sensing image header file. is the band response function of the said band.

[0032] Optionally, obtaining the solar zenith angle, relative azimuth, and observation zenith angle of each pixel in the target remote sensing image includes:

[0033] Interpolating the solar zenith angle, observation zenith angle, and relative azimuth data of the target remote sensing image according to a preset resolution by using the nearest - neighbor method to obtain the solar zenith angle, relative azimuth, and observation zenith angle of each pixel.

[0034] Optionally,

[0035] The first setting range is from 0.001 to 0.8, and the first preset step size is 0.001;

[0036] The second setting range is from 0 degrees to 70 degrees, and the second preset step size is 1 degree;

[0037] The third setting range is from 0 degrees to 180 degrees, and the third preset step size is 10 degrees;

[0038] The observation zenith angle takes the zero points of the Legendre polynomial on the interval (0, 1).

[0039] Another aspect of the present invention provides a general - purpose Rayleigh scattering calculation system for water - color remote sensing, and the system executes the general - purpose Rayleigh scattering calculation method for water - color remote sensing provided in the foregoing aspect.

[0040] The present invention discloses a general - purpose Rayleigh scattering calculation method and system for water - color remote sensing. First, a Rayleigh scattering data set is formed based on observation conditions and atmospheric parameter data, and then this data set is input into a machine - learning method for training to obtain a Rayleigh scattering prediction model. Finally, the observation conditions and atmospheric parameter data of each band and each pixel of the target remote sensing image are obtained and input into the trained prediction model to calculate the Rayleigh scattering of each pixel and each band of the target remote sensing image. The method and system disclosed by the present invention are general - purpose, not specific to a particular sensor, and are applicable to all existing and future water - color remote - sensing sensors. At the same time, they have the advantages of high efficiency and high precision, and avoid the errors brought by the lookup - table method in the linear interpolation process.

[0041] The description of the invention content is provided to introduce the selection of concepts in a simplified form, which will be further described in the following detailed implementation. The description of the invention content is not intended to identify the important features or essential features of the present disclosure, nor is it intended to limit the scope of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The above and other objects, features, and advantages of the present disclosure will become more apparent by describing the exemplary embodiments of the present disclosure in more detail with reference to the accompanying drawings, in which, in the exemplary embodiments of the present disclosure, the same reference numerals generally represent the same components.

[0043] Figure 1 It is a schematic flowchart of a general Rayleigh scattering calculation method for ocean color remote sensing provided by an embodiment of the present invention;

[0044] Figure 2 It is for implementing Figure 1 a schematic flowchart of step S100 in

[0045] Figure 3 It is for implementing Figure 1 a schematic flowchart of step S200 in

[0046] Figure 4 It is for implementing Figure 1 a schematic flowchart of step S300 in Detailed implementation manners

[0047] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0048] The term "including" and its variations used herein mean open inclusion, i.e., "including but not limited to". Unless otherwise specified, the term "or" means "and / or". The term "based on" means "at least partially based on". The terms "an exemplary embodiment" and "an embodiment" mean "at least one exemplary embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc. may refer to different or the same objects. There may be other explicit and implicit definitions hereinafter.

[0049] Figure 1 It is a schematic flowchart of a general Rayleigh scattering calculation method for ocean color remote sensing provided by an embodiment of the present invention, as Figure 1 shown, the method includes the following steps:

[0050] Step S100: Input multiple groups of preset observation conditions and atmospheric parameter data into the radiative transfer model PCOART respectively, and calculate the corresponding Rayleigh scattering Stokes vector data.

[0051] The observation condition data includes at least the solar zenith angle, the relative azimuth, and the observation zenith angle, and the atmospheric parameters include at least the Rayleigh scattering optical depth of atmospheric molecules.

[0052] In an embodiment disclosed by the present invention, as Figure 2 shown, the step S100 can be implemented in the following manner:

[0053] Step S101: Take equidistant values of the Rayleigh scattering optical depth of atmospheric molecules at a first preset step within a first set range to obtain a set of values of the Rayleigh scattering optical depth of atmospheric molecules.

[0054] The optical depth of atmospheric molecules is one of the important parameters affecting Rayleigh scattering, reflecting the scattering and absorption ability of the atmosphere to light. The larger the value, the stronger the scattering effect when light propagates in the atmosphere.

[0055] In a specific embodiment disclosed by the present invention, the first set range is from 0.001 to 0.8, and the first preset step is 0.001, so as to obtain a set of equidistant Rayleigh scattering optical depth data of atmospheric molecules, ensuring coverage from the ultraviolet band to the short-wave infrared band, and avoiding omission or error accumulation in numerical calculations. In addition, the selection of a dense step helps to obtain a smoother result in model calculation and more accurately capture the influence trend of the Rayleigh scattering optical depth of atmospheric molecules on the scattering Stokes vector.

[0056] Step S102: Take equidistant values of the solar zenith angle at a second preset step within a second set range to obtain a set of values of the solar zenith angle.

[0057] The solar zenith angle determines the angle at which sunlight enters the atmosphere, directly affecting the directionality and intensity change of Rayleigh scattering. For a smaller solar zenith angle, sunlight is close to vertical incidence, and the path through the atmosphere is shorter, so the scattering effect is weaker; while for a larger solar zenith angle, the incident angle of sunlight is more inclined, the path through the atmosphere becomes longer, and the scattering effect increases accordingly.

[0058] In a specific embodiment disclosed by the present invention, the second set range is from 0 degrees to 70 degrees, and the second preset step is 1 degree, covering different solar incident angles from near noon to near sunset, and providing a sufficiently dense angle distribution to improve the calculation accuracy.

[0059] Step S103: Take equidistant values of the relative azimuth at a third preset step within a third set range to obtain a set of values of the relative azimuth.

[0060] The relative azimuth describes the angular relationship between the observation direction and the sunlight incident direction, and the polarization characteristics of Rayleigh scattering change significantly at different relative azimuth angles.

[0061] In a specific embodiment disclosed by the present invention, the third setting range is from 0 degree to 180 degrees, and the third preset step is 10 degrees, ensuring that the calculated data covers all possible scattering geometric relationships.

[0062] Step S104: Preset the value of the observation zenith angle.

[0063] In a specific embodiment disclosed by the present invention, the observation zenith angle takes the zero points of the Legendre polynomial in the interval (0, 1).

[0064] Step S105: Based on the value set of the Rayleigh scattering optical thickness of atmospheric molecules, the value set of the solar zenith angle, the value set of the relative azimuth, and the preset observation zenith angle, combine different observation conditions and atmospheric parameters to obtain multiple sets of different observation conditions and atmospheric parameter data.

[0065] In an embodiment disclosed by the present invention, step S105 can be implemented in the following manner:

[0066] Arrange and combine the values in the value set of the Rayleigh scattering optical thickness of atmospheric molecules, the value set of the solar zenith angle, and the value set of the relative azimuth with the value of the observation zenith angle to form multiple sets of observation conditions and atmospheric parameter data.

[0067] In order to obtain complete Rayleigh scattering characteristic data, the above four parameters need to be fully combined, that is, each value of the Rayleigh scattering optical thickness of atmospheric molecules is combined with all values of the solar zenith angle in turn, further combined with all relative azimuth values, and finally combined with all observation zenith angle values to form all possible combinations of observation conditions.

[0068] For example, use nested loops to traverse different parameter values, or adopt matrix operation methods to generate all combined data at once. Through the permutation and combination method, the variation trend of Rayleigh scattering under different conditions can be comprehensively analyzed.

[0069] Step S106: Input each set of observation conditions and atmospheric parameter data into the radiative transfer model PCOART respectively to obtain the corresponding Rayleigh scattering Stokes vector.

[0070] After completing the permutation and combination of different observation conditions and atmospheric parameters, each set of data will be used as input parameters and sent into the PCOART radiative transfer model respectively to calculate the Rayleigh scattering Stokes vector under the corresponding conditions.

[0071] The Stokes vector of Rayleigh scattering consists of the I component (total radiance), the Q component, and the U component (two orthogonal components of the linear polarization degree). Among them, the I component (light intensity) represents the total intensity of the scattered light, that is, the brightness of the light received in the observation direction, which is jointly affected by the atmospheric optical thickness, solar zenith angle, relative azimuth, and observation zenith angle. When the optical thickness is large, the intensity of the scattered light is usually high, and under different solar zenith angles and azimuth angles, the distribution of the light intensity may be different.

[0072] The Q component (the first component of the polarization degree) describes the polarization degree of light in the horizontal-vertical direction. In Rayleigh scattering, when the scattering angle is close to 90°, the linear polarization degree of the scattered light is usually high, and the polarization direction is orthogonal to the scattering plane. Therefore, the value of the Q component will show different positive and negative distributions with the change of the scattering angle, reflecting the polarization direction of the scattered light.

[0073] The U component (the second component of the polarization degree) represents the polarization degree of light in the ±45° direction, and its value depends on the rotation of the observation direction relative to the scattering plane. Together with the Q component, it determines the polarization direction of the scattered light.

[0074] In the actual calculation process, each set of observation conditions and atmospheric parameter data will affect the Stokes vector of Rayleigh scattering, and the PCOART model will calculate the corresponding I, Q, and U component values based on the input parameters. By analyzing the calculation results of a large number of different parameter combinations, the polarization characteristic distribution of Rayleigh scattering under different atmospheric conditions can be obtained.

[0075] Step S200: Train an XGBoost machine learning model based on the observation conditions, atmospheric parameter data, and the corresponding Rayleigh scattering Stokes vector data to obtain a Rayleigh scattering prediction model.

[0076] In an embodiment disclosed in the present invention, as Figure 3 shown, the following method can be used to implement step S200:

[0077] Step S201: Divide all sets of observation conditions, atmospheric parameter data, and the corresponding Rayleigh scattering Stokes vector data into a training set and a test set according to a preset ratio.

[0078] For example, all sets of observation conditions, atmospheric parameter data, and corresponding Stokes vector data can be randomly divided into a training set (80%) and a test set (20%) in a ratio of 8:2. Among them, the training set is used to train the XGBoost model to learn the characteristic relationship of Rayleigh scattering, while the test set is used to evaluate the generalization ability of the trained model on unseen data. This data division method can effectively prevent the model from overfitting and ensure its good prediction performance. When randomly dividing the data, techniques such as cross-validation can be used to ensure the uniformity of the data, enabling the model to learn effective patterns under different conditions.

[0079] Step S202: Use the optical thickness of atmospheric molecular Rayleigh scattering, solar zenith angle, relative azimuth, and observation zenith angle as input features, and use the I component, Q component, and U component of the Rayleigh scattering Stokes vector as output features. Train the XGBoost machine learning model with the training set to obtain the Rayleigh scattering prediction model.

[0080] XGBoost is a machine learning algorithm based on gradient-boosted decision trees (GBDT), with strong non-linear fitting ability, and can handle high-dimensional and non-linear data well. During the training process, the XGBoost model can learn the complex mapping relationship between the input features (optical thickness of atmospheric molecular Rayleigh scattering, solar zenith angle, relative azimuth, observation zenith angle) and the output variables (Stokes vector components I, Q, U), enabling the model to predict the Rayleigh scattering characteristics under new observation conditions.

[0081] Step S203: Test the trained Rayleigh scattering prediction model with the test set.

[0082] Use the test set to evaluate the trained Rayleigh scattering prediction model. For example, by calculating metrics such as mean squared error (MSE), root mean squared error (RMSE), and coefficient of determination (R 2 ), etc., to measure the prediction ability and generalization performance of the model. In addition, visualization means (such as scatter plots, residual analysis) can be used to compare the differences between the true values and the predicted values, so as to determine whether there are systematic biases in the model and make corresponding adjustments, such as optimizing the model hyperparameters, increasing the training data, or adjusting the feature engineering strategy.

[0083] Step S300: Obtain the observation conditions and atmospheric parameter data of the target remote sensing image and input them into the Rayleigh scattering prediction model to obtain the Rayleigh scattering Stokes vector data of each pixel in the target remote sensing image.

[0084] In an embodiment disclosed by the present invention, as Figure 4 shown, step S300 can be implemented in the following manner:

[0085] Step S301: Obtain the Rayleigh scattering optical depth values of atmospheric molecules for each band in the target remote sensing image.

[0086] In the disclosed embodiments of the present invention, taking the remote sensing images collected by Sentinel-2 A / B satellites as an example. The Sentinel-2A / B satellites are equipped with a multi-spectral sensor (MSI), and the response characteristics of each band are different. The equivalent Rayleigh scattering optical depth of atmospheric molecules for each band can be calculated according to the band response function provided by ESA.

[0087] Each band of the remote sensing sensor is not a single wavelength, but covers a certain spectral range, and the wavelength responses within this range are different. Therefore, it is necessary to use the band response function to perform weighted averaging on the Rayleigh scattering optical depth of atmospheric molecules at different wavelengths to obtain the equivalent Rayleigh scattering optical depth of atmospheric molecules for each band. .

[0088] The following formula can be used to calculate the equivalent Rayleigh scattering optical depth value of each band:

[0089]

[0090] where band is the band of the remote sensing sensor, is the wavelength of the band, is the Rayleigh scattering optical depth value of atmospheric molecules for the band, is the solar irradiance at the average sun-earth distance obtained in advance, which is obtained from the remote sensing image header file, is the band response function of the band.

[0091] Step S302: Obtain the solar zenith angle, relative azimuth, and observation zenith angle of each pixel in the target remote sensing image.

[0092] In an embodiment disclosed in the present invention, the nearest neighbor method is used to interpolate the solar zenith angle, observation zenith angle, and relative azimuth data of the target remote sensing image according to a preset resolution to obtain the solar zenith angle, relative azimuth, and observation zenith angle of each pixel.

[0093] In remote sensing image processing, the solar zenith angle, relative azimuth, and observation zenith angle are usually provided by the metadata of the image. In order to obtain the corresponding values of each pixel in the remote sensing image, these angle parameters need to be interpolated according to the spatial resolution of the target remote sensing image.

[0094] In this embodiment, the nearest neighbor interpolation method is used to resample the observation angle data to match the pixel resolution of the remote sensing image. The advantage of the nearest neighbor interpolation method is that it is computationally simple and can maintain the discrete characteristics of the original data, avoiding non-physical smoothing of the angle data. After interpolation, each pixel will have accurate solar zenith angle, relative azimuth, and observation zenith angle information for Rayleigh scattering calculation.

[0095] Step S303: Input the Rayleigh scattering optical depth values of atmospheric molecules in each band and the solar zenith angle, relative azimuth, and observation zenith angle of each pixel into the Rayleigh scattering prediction model to obtain the Rayleigh scattering Stokes vector data corresponding to each pixel in each band of the target remote sensing image.

[0096] After obtaining the observation geometric parameters of each pixel, input them together with the calculated equivalent Rayleigh scattering optical depth of atmospheric molecules into the Rayleigh scattering prediction model (i.e., a machine learning model trained based on XGBoost). For each pixel, under each band of Sentinel-2 A / B, the model will predict its corresponding Rayleigh scattering Stokes vector (I, Q, U).

[0097] An embodiment of the present invention also provides a general water color remote sensing Rayleigh scattering calculation system, which can execute the general water color remote sensing Rayleigh scattering calculation method proposed in the foregoing embodiment.

[0098] The various embodiments of the present disclosure have been described above. The above description is exemplary and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art in the technical field without departing from the scope and spirit of the described embodiments. The selection of the terms used herein is intended to best explain the principles of the embodiments, the practical application, or the technical improvement of the technology in the market, or to enable other ordinary skilled persons in the technical field to understand the embodiments disclosed herein.

Claims

1. A universal Rayleigh scattering calculation method for water color remote sensing, characterized in that: include: Inputting multiple sets of preset observation conditions and atmospheric parameter data into the radiation transfer model PCOART respectively, and calculating corresponding Rayleigh scattering Stokes vector data, wherein the observation condition data at least includes the solar zenith angle, the relative azimuth and the observation zenith angle, and the atmospheric parameter data at least includes the Rayleigh scattering optical thickness of atmospheric molecules; the Rayleigh scattering Stokes vector is composed of an I component, a Q component and a U component; Based on the observation conditions and atmospheric parameter data and the corresponding Rayleigh scattering Stokes vector data, an XGBoost machine learning model is trained to obtain a Rayleigh scattering prediction model, including: All groups of observation conditions and atmospheric parameter data and corresponding Rayleigh scattering Stokes vector data are divided into training sets and test sets according to the preset ratio; The Rayleigh scattering optical thickness of atmospheric molecules, solar zenith angle, relative azimuth and observation zenith angle are used as input features, and the I component, Q component and U component of the Rayleigh scattering Stokes vector are used as output features. The training set is used to train the XGBoost machine learning model to obtain the Rayleigh scattering prediction model. The trained Rayleigh scattering prediction model is tested using the test set; The observation conditions and atmospheric parameter data of the target remote sensing image are obtained and input into the Rayleigh scattering prediction model to obtain the Rayleigh scattering Stokes vector data of each pixel in the target remote sensing image.

2. The calculation method according to claim 1, characterized in that: The method of inputting a plurality of preset observation conditions and atmospheric parameter data into the radiation transfer model PCOART to calculate the corresponding Rayleigh scattering Stokes vector data includes: Taking values ​​of the atmospheric molecule Rayleigh scattering optical thickness values ​​in equal steps with a first preset step length within a first set range to obtain a set of atmospheric molecule Rayleigh scattering optical thickness values; Taking values ​​of the solar zenith angle in equal steps with a second preset step length within a second set range to obtain a solar zenith angle value set; Taking values ​​of the relative azimuth in equal steps with a third preset step length within a third setting range to obtain a relative azimuth value set; Preset the value of the observation zenith angle; Based on the atmospheric molecule Rayleigh scattering optical thickness value set, the solar zenith angle value set, the relative azimuth value set and the preset observation zenith angle, different observation conditions and atmospheric parameters are combined to obtain multiple sets of different observation conditions and atmospheric parameter data; Each set of observation conditions and atmospheric parameter data is input into the radiation transfer model PCOART to obtain the corresponding Rayleigh scattering Stokes vector.

3. The calculation method according to claim 2, characterized in that: The method combines different observation conditions and atmospheric parameters based on the atmospheric molecule Rayleigh scattering optical thickness value set, the solar zenith angle value set, the relative azimuth value set and the preset observation zenith angle to obtain multiple groups of different observation conditions and atmospheric parameter data, including: The values ​​in the atmospheric molecule Rayleigh scattering optical thickness value set, the solar zenith angle value set and the relative azimuth value set are arranged and combined with the value of the observation zenith angle to form multiple sets of observation conditions and atmospheric parameter data.

4. The calculation method according to claim 1, characterized in that: The acquisition of observation conditions and atmospheric parameter data of the target remote sensing image and inputting them into the Rayleigh scattering prediction model to obtain Rayleigh scattering Stokes vector data of each pixel in the target remote sensing image includes: Obtain the Rayleigh scattering optical thickness values ​​of atmospheric molecules in each band in the target remote sensing image; Obtain the solar zenith angle, relative azimuth and observation zenith angle of each pixel in the target remote sensing image; The Rayleigh scattering optical thickness values ​​of atmospheric molecules in each band and the solar zenith angle, relative azimuth and observation zenith angle of each pixel are input into the Rayleigh scattering prediction model to obtain the Rayleigh scattering Stokes vector data corresponding to each pixel in the target remote sensing image in each band.

5. The calculation method according to claim 4, characterized in that: The step of obtaining the atmospheric molecule Rayleigh scattering optical thickness values ​​of each band in the target remote sensing image includes: The equivalent atmospheric molecule Rayleigh scattering optical thickness value of each band is calculated using the following formula: Where, band is the band of the remote sensing sensor, λ is the wavelength of the band, τ r is the Rayleigh scattering optical thickness value of atmospheric molecules in the band, F0 is the solar irradiance at the average sun-earth distance obtained in advance, obtained from the remote sensing image header file, S band is the band response function of the band.

6. The calculation method according to claim 4, characterized in that: The step of obtaining the solar zenith angle, relative azimuth and observation zenith angle of each pixel in the target remote sensing image comprises: The nearest neighbor method is used to interpolate the solar zenith angle, observation zenith angle and relative azimuth data of the target remote sensing image according to the preset resolution to obtain the solar zenith angle, relative azimuth and observation zenith angle of each pixel.

7. The calculation method according to claim 2, characterized in that: The first setting range is from 0.001 to 0.8, and the first preset step is 0.001; The second setting range is from 0 degrees to 70 degrees, and the second preset step is 1 degree; The third setting range is from 0 degrees to 180 degrees, and the third preset step is 10 degrees; The observed zenith angle is taken as the zero point of the Legendre polynomial in the interval (0,1).

8. A universal Rayleigh scattering calculation system for water color remote sensing, characterized in that: The system executes the method according to any one of claims 1-7.

Images