Cooperative inversion method and system for atmosphere and ocean parameters of offshore area

Abstract

The invention provides an offshore area atmosphere and ocean parameter collaborative inversion method and system, and relates to the technical field of satellite remote sensing, and the method comprises the steps: obtaining to-be-inverted parameters which comprise an atmosphere parameter and an ocean parameter, calling a multi-parameter physical information neural network model to carry out the ocean water-leaving radiation calculation of the ocean parameter, and calculating the ocean water-leaving radiation of the ocean parameter; an atmospheric aerosol model is called to carry out parameterization calculation on atmospheric parameters, and scattering parameters and absorption coefficients of atmospheric aerosol are obtained; inputting the ocean water-leaving radiation quantity and the scattering parameter and the absorption coefficient of the atmospheric aerosol into a forward radiation transfer model, and carrying out radiation transfer calculation to obtain entrance pupil radiation brightness of the top of the atmospheric layer; and finally, performing parameter inversion by combining a satellite measurement radiation signal to obtain an inversion parameter corresponding to the to-be-inverted parameter. According to the invention, problems existing in an atmospheric correction algorithm for ocean water color remote sensing in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of satellite remote sensing, and in particular to a method and system for simultaneously retrieving atmospheric and oceanic parameters in coastal waters. BACKGROUND

[0002] Atmosphere-ocean system research has important practical value in climate change, ecological environment monitoring, etc. Remote sensing of atmospheric and oceanic parameters using satellite remote sensing technology is currently the most effective method. The radiation at the top of the atmosphere is the signal observed by the satellite, which includes atmospheric and oceanic parts. The scattering signal of Rayleigh scattering of atmospheric molecules and aerosols accounts for the majority, and the reflection and off-water radiation of the ocean surface accounts for only 10-15% of the zenith signal observed by the satellite. The chlorophyll content, colored soluble matter, suspended particles and other optical characteristic parameters of the water quality of the ocean can be retrieved from the off-water radiation.

[0003] In the prior art, the atmospheric correction algorithm for ocean color remote sensing generally assumes that the off-water radiation is in the near-infrared or short-wave infrared (referred to as reference wavelength), estimates the aerosol scattering signal at the reference wavelength, and then extrapolates the aerosol scattering signal at visible or near-infrared wavelengths according to the aerosol model spectrum to obtain the off-water radiance at these wavelengths, thereby achieving atmospheric correction. However, there are still some problems. Firstly, the extrapolation error will significantly increase with the increase of the distance of the extrapolated spectrum, leading to the accumulation and amplification of atmospheric correction errors in the short wave. Secondly, the widely existing absorbing aerosols in the coastal area cannot be effectively handled, resulting in a significant "negative value" problem in the atmospheric correction results at blue wavelengths. SUMMARY

[0004] The present application provides a method and system for simultaneously retrieving atmospheric and oceanic parameters in coastal waters to solve the problems existing in the atmospheric correction algorithm for ocean color remote sensing in the prior art.

[0005] The present application provides a method for simultaneously retrieving atmospheric and oceanic parameters in coastal waters, comprising: obtaining the parameters to be retrieved, including atmospheric parameters and oceanic parameters, the oceanic parameters including chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter at a wavelength of 440 nanometers, and backscattering coefficient of particulate matter at a wavelength of 660 nanometers, and the atmospheric parameters including effective radius, effective variance, particle concentration, component information and complex refractive index of atmospheric aerosols; calling a multi-parameter physical information neural network model to calculate the ocean off-water radiation of the oceanic parameters to obtain the ocean off-water radiation, and calling an atmospheric aerosol model to perform parameterization calculation on the atmospheric parameters to obtain the scattering parameters and absorption coefficient of atmospheric aerosols; inputting the marine off-water radiation, the scattering parameter and the absorption coefficient of the atmospheric aerosol into a forward radiation transfer model, performing radiation transfer calculation to obtain an entrance pupil radiation brightness at the top of the atmosphere layer; performing parameter inversion according to a satellite measured radiation signal and the entrance pupil radiation brightness to obtain an inversion parameter corresponding to the to-be-inverted parameter.

[0006] In some embodiments, the multi-parameter physical information neural network model is obtained by training a training sample set, and a construction process of the training sample set comprises: constructing a marine parameter sample; performing normalization processing on the marine parameter sample to obtain input data of the training sample set; calling a preset marine bio-optical model to perform marine off-water radiation calculation on the marine parameter sample to obtain a data sample of marine off-water radiation; performing Fourier expansion calculation on the data sample at a relative azimuth angle at each wavelength in a light wavelength range to obtain a Fourier coefficient corresponding to the wavelength, the relative azimuth angle being a difference between a solar azimuth angle and a satellite azimuth angle, and the light wavelength range being 250 nanometers to 800 nanometers with a wavelength interval of 5 nanometers; performing dimension reduction processing on the Fourier coefficient by principal component analysis, and taking the Fourier coefficient corresponding to a principal component component obtained by the dimension reduction processing as output data of the training sample set.

[0007] In some embodiments, the process of the Fourier coefficient dimension reduction processing comprises: performing principal component analysis on the Fourier coefficient of each wavelength to obtain a candidate principal component component; performing data reconstruction on a feature vector calculated based on the candidate principal component component to obtain reconstructed data; calculating a relative error percentage at each wavelength according to the Fourier coefficient and the reconstructed data; when an average value of the relative error percentage is less than 0.1%, taking the candidate principal component component as a principal component component after the dimension reduction processing.

[0008] In some embodiments, in the marine parameter sample, an absorption coefficient of particulate matter under light with a wavelength of 440 nanometers satisfies the following constraint relationship: a backscattering coefficient of particulate matter under light with a wavelength of 660 nanometers satisfies the following constraint relationship: wherein, a represents the absorption coefficient of particulate matter under light with a wavelength of 440 nanometers, represents the absorption coefficient of phytoplankton at a wavelength of 440 nm. It represents the backscattering coefficient of particles under the light with a wavelength of 660 nanometers. represents the absorption coefficient of phytoplankton at a wavelength of 660 nanometers. It represents the absorption coefficient of particles at a wavelength of 660 nanometers.

[0009] In some embodiments, the training process of the multi-parameter physical information neural network model includes: Constructing a root mean square error loss function based on the input data of the training sample set and the output data of the training sample set; Back propagation is performed in the multi-parameter physical information neural network model according to the root mean square error loss function to update the parameters of the multi-parameter physical information neural network model.

[0010] In some embodiments, performing parameter inversion based on the satellite-measured radiation signal and the entrance pupil radiation brightness to obtain inversion parameters corresponding to the parameters to be inverted includes: Determining an error value between the satellite measured radiation signal and the entrance pupil radiation brightness; Constructing an inversion cost function for the state vector of the parameter to be inverted according to the error value; The minimum value of the inversion cost function is calculated, and the state vector of the parameter to be inverted when the inversion cost function obtains the minimum value is determined as the inversion parameter corresponding to the parameter to be inverted.

[0011] The present invention also provides a nearshore sea area atmospheric and ocean parameter collaborative inversion system, comprising: An acquisition module is used to acquire parameters to be inverted, wherein the parameters to be inverted include atmospheric parameters and ocean parameters. The ocean parameters include chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter at a wavelength of 440 nanometers, and backscattering coefficient of particulate matter at a wavelength of 660 nanometers. The atmospheric parameters include effective radius, effective variance, particle number concentration, composition information, and complex refractive index of atmospheric aerosols. a calculation module, configured to call a multi-parameter physical information neural network model to perform ocean water-leaving radiation calculation on the ocean parameters to obtain ocean water-leaving radiation, and call an atmospheric aerosol model to perform parameterized calculation on the atmospheric parameters to obtain scattering parameters and absorption coefficients of atmospheric aerosols; a transmission module, configured to input the ocean water-leaving radiation, the scattering parameters and the absorption coefficient of the atmospheric aerosol into a forward radiation transfer model, perform radiation transfer calculations, and obtain the entrance pupil radiation brightness at the top of the atmosphere; An inversion module is configured to perform parameter inversion according to the satellite measured radiation signal and the entrance pupil radiation brightness to obtain an inversion parameter corresponding to the to-be-inverted parameter The application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method for simultaneously inverting atmospheric and marine parameters in a near-shore sea area according to any one of the above.

[0012] The application further provides a non-transitory computer-readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the method for simultaneously inverting atmospheric and marine parameters in a near-shore sea area according to any one of the above.

[0013] The application further provides a computer program product, which includes a computer program, and the computer program is executable on a processor to implement the method for simultaneously inverting atmospheric and marine parameters in a near-shore sea area according to any one of the above.

[0014] The application provides the method and system for simultaneously inverting atmospheric and marine parameters in a near-shore sea area, which utilizes a multi-parameter physical information neural network model to realize synchronous inversion of atmospheric and marine parameters on the basis of atmospheric parameters and marine parameters containing an observation angle, overcomes the limitation of an atmospheric correction algorithm of marine water color remote sensing, and improves the efficiency of business calculation through the neural network algorithm, the calculation result has certain interpretability, is beneficial to coupling of atmospheric data and a marine system, and can be applied to various satellite remote sensing data scenes of a polarization observation angle. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description one by one. Obviously, the drawings in the following description are some embodiments of the application, and those skilled in the art can also obtain other drawings according to these drawings without any creative effort.

[0016] Figure 1 is a flowchart of the method for simultaneously inverting atmospheric and marine parameters in a near-shore sea area provided by the application.

[0017] Figure 2 is a schematic diagram of the framework principle of the method for simultaneously inverting atmospheric and marine parameters in a near-shore sea area provided by the application.

[0018] Figure 3 is a training schematic diagram of the multi-parameter physical information neural network model provided by the application.

[0019] Figure 4 is an application schematic diagram of the forward radiation transfer model provided by the application.

[0020] Figure 5 It is a structural schematic diagram of the collaborative inversion system for atmospheric and oceanic parameters in nearshore waters provided by the present invention.

[0021] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0023] The following describes the method and system for collaborative inversion of nearshore atmospheric and oceanic parameters of the present invention with reference to the accompanying drawings. Figure 1 This is a flow chart of the collaborative inversion method for atmospheric and oceanic parameters in nearshore waters provided by the present invention. Figure 1 As shown, the method includes the following steps 101 to 104, which are described one by one below.

[0024] Step 101: Obtain parameters to be inverted.

[0025] First, obtain the parameters to be inverted, including atmospheric parameters and ocean parameters, which are measured by various measuring instruments, satellite observations or further calculations in the nearshore waters. Among them, ocean parameters include chlorophyll concentration , solar zenith angle , satellite observation zenith angle , the absorption coefficient of particles at a wavelength of 440 nanometers and the backscattering coefficient of particles at a wavelength of 660 nanometers It should be noted that chlorophyll concentration refers to the concentration of chlorophyll a. The specific introduction of ocean parameters is shown in Table 1 below: Table 1:

[0026] The atmospheric parameters are mainly atmospheric aerosol data, including the effective radius, effective variance, particle number concentration, composition information and complex refractive index of atmospheric aerosols. The complex refractive index is a complex number, including real and imaginary parts.

[0027] Step 102, calling the multi-parameter physical information neural network model to calculate the ocean off-water radiation of the marine parameters to obtain the ocean off-water radiation, and calling the atmospheric aerosol model to parameterize the atmospheric parameters to obtain the scattering parameters and absorption coefficients of the atmospheric aerosols.

[0028] In the embodiment of the present application, as shown in Figure 2 , the parameter inversion process includes three stages, the first stage is to calculate the ocean off-water radiation of the marine parameters to the radiation transfer process in the coastal sea area by using the multi-parameter physical information neural network model. Therefore, the multi-parameter physical information neural network model is called here to calculate the ocean off-water radiation of the marine parameters to obtain the ocean off-water radiation. The multi-parameter physical information neural network model needs to be trained in advance, and here the multi-parameter physical information neural network model is trained by the training sample set. The construction process of the training sample set is described below in combination with Figure 3 .

[0029] First, the marine parameter sample needs to be constructed, which can be randomly selected from the marine parameters collected in step 101 as the marine parameter sample, that is, the marine parameters collected in the coastal sea area, which are affected by the constraints of the ground-based observation information, including chlorophyll concentration , the absorption coefficient of particulate matter under light with a wavelength of 440 nanometers , the backscattering coefficient of particulate matter under light with a wavelength of 660 nanometers , and the geometric parameters of satellite observation, including the solar zenith angle , the satellite observation zenith angle , in addition to the relative azimuth angle , that is the difference between the solar azimuth angle and the satellite azimuth angle .

[0030] In the following embodiment, since the marine parameters in the marine ground-based observation data need to satisfy the actual physical constraint relationship, the marine parameter sample needs to be judged to satisfy the physical constraint relationship, so as to screen the marine parameter sample, that is, to screen out some marine parameter samples that do not satisfy the physical constraint relationship.

[0031] In the marine parameter sample, the absorption coefficient of particulate matter under light with a wavelength of 440 nanometers satisfies the following constraint relationship: (1) And the backscattering coefficient of particulate matter under light with a wavelength of 660 nanometers satisfies the following constraint relationship: (2) Wherein, an absorption coefficient of particulate matter at a wavelength of 440 nm, an absorption coefficient of phytoplankton at a wavelength of 440 nm, a backscattering coefficient of particulate matter at a wavelength of 660 nm, an absorption coefficient of phytoplankton at a wavelength of 660 nm, an absorption coefficient of particulate matter at a wavelength of 660 nm.

[0032] Through the constraint relationship in the above formula (1) and formula (2), the data of the constructed marine parameter sample can be filtered, so that each marine parameter sample has practical significance, ensures the effectiveness of the observation data, improves the quality of the data sample, and does not affect the performance of the multi-parameter physical information neural network model.

[0033] The constructed marine parameter sample is subjected to data normalization processing, as shown in Table 1 above, there are many types of marine parameters and different data distributions, so data unification is needed, and therefore the marine parameter sample needs to be normalized to obtain the input data of the training sample set. The conversion formula of the normalization processing is as follows: (3) In the above formula (3), a and b are weight parameters used for normalization, y is the input data of the training sample set, i.e. the normalized marine parameter sample, and the value range is -1 to 1. x is the marine parameter sample that needs to be normalized. For the marine parameter sample with a lognormal distribution, the logarithm is taken first and then normalized as x. For the uniformly distributed marine parameter sample, x is directly normalized.

[0034] In addition, the training sample set also needs output data, which is obtained from the marine water-leaving radiance calculated from these marine parameter samples. Here, the marine water-leaving radiation of the marine parameter sample can be directly calculated by calling a preset marine bio-optical model to obtain the data sample of the marine water-leaving radiation. The marine bio-optical model can comprehensively analyze the coupling between the optical properties of the water body and the radiation transmission process, but since the model cannot combine multiple marine parameters to calculate the marine water-leaving radiation, the accuracy of the calculation result is poor, and therefore it can only be used for preliminary calculation. The marine bio-optical model can be a Morel bio-optical model, a Garver-Siegel-Maritorena model, or a specific trained machine learning model.

[0035] The calculated data sample is then Fourier expanded at the relative azimuth angle for each wavelength to obtain a plurality of Fourier coefficients at each wavelength. Here, the data sample is Fourier expanded at the relative azimuth angle for each wavelength in the light wavelength range to obtain the Fourier coefficients of the corresponding wavelength. The relative azimuth angle is the difference between the solar azimuth angle and the satellite azimuth angle, the light wavelength range is 250 nanometers to 800 nanometers, the wavelength interval is 5 nanometers, and thus 91 wavelengths can be obtained, and 8 Fourier coefficients can be obtained at each wavelength through Fourier expansion calculation. The formula of the Fourier expansion calculation is as follows: (4) In the above formula (4), are all Fourier expansion coefficients, is the cosine value of the observation zenith angle, is the solar azimuth angle, is the satellite azimuth angle, is the relative azimuth angle, and m represents the number of all Fourier components, is the Dirac function, are the upward and downward components of radiance, respectively, represents the calculated Fourier coefficients, and z represents the altitude.

[0036] Through the above formula (2), a total of 91x8=728 Fourier coefficients can be calculated, and the data dimension is too large, so principal component analysis is performed next to reduce the data dimension, that is, the Fourier coefficients corresponding to the principal components obtained through principal component analysis are used as the output data of the training sample set. The number of principal components can be determined through principal component analysis, the data amount is reduced without losing characteristic information, and the Fourier coefficients corresponding to the finally obtained principal components are used as the output data of the training sample set.

[0037] In the embodiment of the present application, when constructing the training sample set, the data sample of the marine exo-atmospheric radiation calculated according to the marine biological optical model is calculated in the form of Fourier series expansion to obtain the characteristics of all wavelengths, and then principal component analysis is performed to realize dimension reduction. The trained multi-parameter physical information neural network model has strong algorithm portability, can adapt to high-spectral load data, and realizes data dimension reduction through principal component analysis, and also reduces the calculation amount of model training.

[0038] In some embodiments, the Fourier coefficients are reduced in dimension through principal component analysis, which can be realized through the following process. The process of reducing the Fourier coefficients in dimension is specifically introduced as follows.

[0039] First, perform principal component analysis on the Fourier coefficients of each wavelength to obtain candidate principal components. This analysis can be performed using a principal component analysis algorithm. The principal component analysis process outputs specific principal components and the corresponding calculated eigenvectors.

[0040] The eigenvectors calculated from the candidate principal component components are then reconstructed to obtain the reconstructed data. Here, the original data is reconstructed based on the eigenvectors. The principal component components are linear combinations of the input Fourier coefficient data features. Each principal component component corresponds to an eigenvector component, and the Fourier coefficients can be reconstructed and restored based on these eigenvector components.

[0041] Next, the reconstructed Fourier coefficients can be compared with the original input Fourier coefficients to calculate the difference between the two. Here, the relative error percentage is calculated at each wavelength based on the Fourier coefficients and the reconstructed data, and the average relative error percentage for all wavelengths is further calculated.

[0042] Finally, a judgment is made. When the average value of the relative error percentage is less than 0.1%, the candidate principal component is used as the principal component after dimensionality reduction.

[0043] Here, when the average relative error percentage is less than 0.1%, it indicates that the candidate principal components obtained by principal component analysis are accurate, thus achieving dimensionality reduction of the Fourier coefficients. If the average relative error percentage is not less than 0.1%, the principal component analysis needs to be repeated to ensure that the average relative error percentage is less than 0.1%.

[0044] In the embodiment of the present invention, the wavelength range of light involves 91 wavelengths, each wavelength involves 8 Fourier coefficients, and the final data amount for principal component analysis is 728 Fourier coefficients. After principal component analysis, 300 principal component components are obtained. Finally, the 728 Fourier coefficients are reduced to 300 Fourier coefficients, achieving effective compression of redundant data and reducing the training difficulty of the multi-parameter physical information neural network model.

[0045] like Figure 3 As shown, when the multi-parameter physical information neural network model is trained, a root mean square error loss function is constructed based on the input data of the training sample set and the output data of the training sample set.

[0046] Here, the input data of the training sample set is used as the input data of the multi-parameter physical information neural network, and the output data of the training sample set is used as the output data of the multi-parameter physical information neural network. Then, the root mean square error of the input data and the output data can be calculated to construct the root mean square error loss function.

[0047] In the iterative training, the multi-parameter physical information neural network model is back propagated according to the root mean square error loss function to update the parameters of the multi-parameter physical information neural network model. In the back propagation, the gradient of the model is calculated, and the gradient is optimized by the gradient descent algorithm, so as to update the parameters of the multi-parameter physical information neural network model. When the number of training iterations reaches the preset number of rounds or the root mean square error loss function starts to converge, the training is stopped.

[0048] For the marine parameters in the parameters to be inverted, the marine off-water radiation is calculated by the preset marine bio-optical model, and then the calculated result is directly input into the trained multi-parameter physical information neural network model for prediction, and the model finally outputs the final marine off-water radiation.

[0049] For the atmospheric parameters in the parameters to be inverted, the atmospheric aerosol model is called to parameterize and calculate the atmospheric parameters to obtain the scattering parameters and absorption coefficients of the atmospheric aerosol. The atmospheric aerosol model is a model constructed according to the Mie theory and the T-matrix theory, which is respectively used to process spherical particle aerosol and non-spherical particle aerosol. The atmospheric parameters include the effective radius, effective variance, particle number concentration, composition information and complex refractive index of the atmospheric aerosol. Therefore, for the atmospheric parameters of the spherical particle aerosol, the scattering calculation is performed by the Mie theory to obtain the scattering coefficient and absorption coefficient of the corresponding aerosol, and for the atmospheric parameters of the non-spherical particle aerosol, the scattering calculation is performed by the T-matrix theory to obtain the scattering coefficient and absorption coefficient of the corresponding aerosol.

[0050] In step 103, the marine off-water radiation, the scattering parameters and the absorption coefficients of the atmospheric aerosol are input into the forward radiation transfer model to perform radiation transfer calculation to obtain the entrance pupil radiation brightness at the top of the atmospheric layer.

[0051] As shown in Figure 2 In the second stage of the parameter inversion process, the forward radiation transfer model is used to realize the coupling of the atmosphere and the ocean, to perform radiation transfer between the physical and chemical parameters of the atmospheric aerosol and the marine off-water radiation, to calculate the radiance information observed by the satellite as the simulation value of the satellite observation radiation signal.

[0052] Referring to Figure 4The output of the atmospheric aerosol model (i.e., the scattering and absorption coefficients of aerosols) and the ocean water-leaving radiation calculated by the multi-parameter physical information neural network model are directly input into the forward radiative transfer model. The forward radiative transfer model requires setting satellite observation geometry parameters, including the number of bands, angles, and whether polarization observation is available. This simulates the entrance pupil radiance brightness at the top of the atmosphere obtained when observing using satellite observation geometry parameters in real-world scenarios. The forward radiative transfer model combines the physical and chemical properties of aerosols with ocean water-leaving radiation to analyze the mutual scattering between the atmosphere and the ocean, including single and multiple scattering.

[0053] The satellite-observed radiation signal is divided into two parts. The first part uses the forward radiation transfer model to calculate atmospheric transmittance based on the satellite observation zenith angle and the scattering and absorption coefficients of atmospheric aerosols. The ocean-leaving radiation is then weighted (multiplied) by the atmospheric transmittance to obtain the ocean radiation. The second part uses the scattering and absorption coefficients of atmospheric aerosols to calculate the path radiation, which can be calculated using the scattering coefficients of atmospheric aerosols and the scattering phase function. The ocean radiation and path radiation are then added together, and the result is nonlinearly fitted using the forward radiation transfer model to obtain the satellite-observed radiation signal, which is the entrance pupil radiance at the top of the atmosphere.

[0054] Step 104: perform parameter inversion based on the satellite-measured radiation signal and the entrance pupil radiation brightness to obtain inversion parameters corresponding to the parameters to be inverted.

[0055] Here, it is necessary to first obtain the satellite measurement radiation signal, that is, the actual satellite observation value. These actual observation values ​​are also obtained under the corresponding number of bands, angles, and polarization observation conditions. The number of bands, angles, and polarization observation conditions set in the forward radiative transfer model are the same. Only in this way can the difference between the actual value and the simulated value be compared. The entrance pupil radiance brightness at the top of the atmosphere is the result of the forward radiative transfer model simulating satellite observations. The difference between the actual value and the simulated value can be used for parameter inversion.

[0056] like Figure 2 As shown in the figure, in the third stage of parameter inversion, the inversion model is called to realize the inversion of ocean parameters and atmospheric parameters based on the entrance pupil radiation brightness of the top of the atmosphere output by the forward radiation transfer model.

[0057] In an embodiment of the present invention, the inversion model can be expressed as the following formula: (5) In the above formula (5), represents the satellite measurement radiation signal obtained, It represents the difference between the satellite measurement radiation signal and the entrance pupil radiation brightness at the top of the atmosphere. represents the processing function of the forward radiative transfer model, is the state vector of the parameters to be inverted.

[0058] The following will specifically introduce the parameter inversion based on the satellite measured radiation signal and the entrance pupil radiation brightness to obtain the inversion parameters corresponding to the parameters to be inverted. The parameter inversion in the embodiment of the present invention adopts a multi-parameter inversion theoretical framework based on optimization theory.

[0059] First, determine the error between the satellite-measured radiation signal and the entrance pupil radiation brightness. This error is a positive number, that is, directly take the absolute value of the difference between the two, which is the value in the above formula (5). Substitute the satellite measured radiation signal into the above formula (5) , so the above formula (5) only has the state vector of the parameter to be inverted An unknown quantity.

[0060] During the inversion calculation, the inversion cost function of the state vector of the parameter to be inverted is constructed based on the error value, which is expressed as: (6) In the above formula (6), F(x) represents the entrance pupil radiance at the top of the atmosphere output by the forward radiation transfer model, represents the processing function of the forward radiation transfer model, x represents the state vector of the parameter to be inverted, y is the input satellite measurement radiation signal, express and The error value of , Represents the error covariance matrix of satellite observations, which is constructed based on the error values. represents the regularization parameter, W represents the weight matrix, Indicates the offset of the state vector x of the parameter to be inverted during the inversion process.

[0061] Based on the optimization inversion theory framework, the entire inversion process can be regarded as a cost function By calculating the minimum value of the inversion cost function, we can make and The error value is the smallest, that is Minimum, so the accuracy of parameter inversion is optimal.

[0062] Finally, the state vector of the parameter to be inverted when the inversion cost function obtains the minimum value is determined as the inversion parameter corresponding to the parameter to be inverted.

[0063] When the above formula (6) When the minimum value is obtained, an x ​​must be determined, and the x here can be used as the , so from the calculation process of formula (6), The parameter inversion process can be expressed as: (7) The parameter explanations in the above formula (7) can be referred to in the above formula (6), and will not be repeated here. Includes inverted atmospheric and ocean parameters.

[0064] In order to verify the inversion performance, the embodiment of the present invention also designs a simulation analysis of multi-angle polarization observation data. That is, in the forward radiation transmission model, by continuously setting the parameters such as the number of satellite observation bands, the number of angles, and whether polarization observation is available, 100 combinations of the state vectors of the parameters to be inverted can be obtained, so that the entrance pupil radiation brightness of 100 tops of the atmosphere can be fitted. Through the inversion method of the above formula (5) and formula (7), the inversion results of 100 state vectors of the parameters to be inverted can be obtained. This verifies the influence of factors such as the number of satellite observation bands, the number of angles, and whether polarization observation is available on parameter inversion.

[0065] This embodiment of the present invention utilizes a multi-parameter physical information neural network model to achieve simultaneous inversion of atmospheric and oceanic parameters based on atmospheric parameters and oceanic parameters including observation angles. This overcomes the limitations of atmospheric correction algorithms for ocean color remote sensing and resolves the inconsistency issues that arise when inverting aerosol parameters first and then performing atmospheric correction. Furthermore, the neural network algorithm improves the efficiency of operational calculations, and its results offer a degree of interpretability, facilitating the coupling of atmospheric data and ocean systems. Furthermore, it can be applied to satellite remote sensing data scenarios with various polarization observation angles.

[0066] The collaborative inversion system for atmospheric and oceanic parameters in nearshore waters provided by the present invention is described below. The collaborative inversion system for atmospheric and oceanic parameters in nearshore waters described below and the collaborative inversion method for atmospheric and oceanic parameters in nearshore waters described above can be referenced to each other.

[0067] like Figure 5As shown, the offshore atmospheric and marine parameter collaborative inversion system provided by the present application specifically comprises the following modules: an acquisition module 501, a calculation module 502, a transmission module 503, and an inversion module 504. The acquisition module 501 is configured to acquire to-be-inverted parameters, wherein the to-be-inverted parameters include atmospheric parameters and marine parameters, the marine parameters include chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter under light with a wavelength of 440 nanometers, and backscattering coefficient of particulate matter under light with a wavelength of 660 nanometers, and the atmospheric parameters include effective radius, effective variance, particle number concentration, component information, and complex refractive index of atmospheric aerosol; the calculation module 502 is configured to call a multi-parameter physical information neural network model to perform marine off-water radiation calculation on the marine parameters, to obtain marine off-water radiation, and to call an atmospheric aerosol model to perform parameterization calculation on the atmospheric parameters, to obtain scattering parameters and absorption coefficients of atmospheric aerosol; the transmission module 503 is configured to input the marine off-water radiation, the scattering parameters and the absorption coefficients of atmospheric aerosol into a forward radiation transmission model, to perform radiation transmission calculation, and to obtain top-of-atmosphere entrance pupil radiance; and the inversion module 504 is configured to perform parameter inversion according to satellite measurement radiation signals and the top-of-atmosphere entrance pupil radiance, to obtain inversion parameters corresponding to the to-be-inverted parameters. It should be noted that the beneficial effects of the offshore atmospheric and marine parameter collaborative inversion system and the offshore atmospheric and marine parameter collaborative inversion method described above can correspond to each other, and therefore the beneficial effects of the offshore atmospheric and marine parameter collaborative inversion system will not be described here.

[0068] Figure 6 An example of an entity structure diagram of an electronic device is shown in FIG. 1. Figure 6As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communications bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communications bus 640. The processor 610 may call the logic instructions in the memory 630 to execute a collaborative inversion method for atmospheric and oceanic parameters in nearshore waters, the method comprising: obtaining parameters to be inverted, the parameters to be inverted comprising atmospheric parameters and oceanic parameters, the oceanic parameters comprising chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter at a wavelength of 440 nanometers, and backscattering coefficient of particulate matter at a wavelength of 660 nanometers, the atmospheric parameters comprising the effective radius, effective variance, particle number concentration, composition information, and complex refractive index of atmospheric aerosols. ; Call the multi-parameter physical information neural network model to perform ocean water-leaving radiation calculation on the ocean parameters to obtain the ocean water-leaving radiation amount, call the atmospheric aerosol model to perform parameterized calculation on the atmospheric parameters to obtain the scattering parameters and absorption coefficients of the atmospheric aerosols; input the ocean water-leaving radiation amount, the scattering parameters and absorption coefficients of the atmospheric aerosols into the forward radiation transfer model, perform radiation transfer calculation, and obtain the entrance pupil radiation brightness at the top of the atmosphere; perform parameter inversion based on the satellite measurement radiation signal and the entrance pupil radiation brightness to obtain the inversion parameters corresponding to the parameters to be inverted.

[0069] Furthermore, the logic instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0070] In another aspect, the present application also provides a computer program product comprising a computer program, which can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, a computer can perform the method for jointly retrieving atmospheric and oceanic parameters in a coastal sea area, which comprises: obtaining parameters to be retrieved, wherein the parameters to be retrieved comprise atmospheric parameters and oceanic parameters, the oceanic parameters comprise chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter at a wavelength of 440 nm, and backscattering coefficient of particulate matter at a wavelength of 660 nm, and the atmospheric parameters comprise effective radius, effective variance, particle number concentration, component information, and complex refractive index of atmospheric aerosols; calling a multi-parameter physical information neural network model to calculate oceanic out-of-water radiation of the oceanic parameters to obtain oceanic out-of-water radiation, calling an atmospheric aerosol model to perform parameterization calculation on the atmospheric parameters to obtain scattering parameters and absorption coefficients of atmospheric aerosols; inputting the oceanic out-of-water radiation, the scattering parameters and the absorption coefficients of the atmospheric aerosols into a forward radiation transfer model to perform radiation transfer calculation to obtain an entrance pupil radiation brightness at the top of the atmosphere; and performing parameter retrieval according to a satellite measured radiation signal and the entrance pupil radiation brightness to obtain retrieved parameters corresponding to the parameters to be retrieved.

[0071] In another aspect, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the computer program implements the method for jointly retrieving atmospheric and oceanic parameters in a coastal sea area, which comprises: obtaining parameters to be retrieved, wherein the parameters to be retrieved comprise atmospheric parameters and oceanic parameters, the oceanic parameters comprise chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter at a wavelength of 440 nm, and backscattering coefficient of particulate matter at a wavelength of 660 nm, and the atmospheric parameters comprise effective radius, effective variance, particle number concentration, component information, and complex refractive index of atmospheric aerosols; calling a multi-parameter physical information neural network model to calculate oceanic out-of-water radiation of the oceanic parameters to obtain oceanic out-of-water radiation, calling an atmospheric aerosol model to perform parameterization calculation on the atmospheric parameters to obtain scattering parameters and absorption coefficients of atmospheric aerosols; inputting the oceanic out-of-water radiation, the scattering parameters and the absorption coefficients of the atmospheric aerosols into a forward radiation transfer model to perform radiation transfer calculation to obtain an entrance pupil radiation brightness at the top of the atmosphere; and performing parameter retrieval according to a satellite measured radiation signal and the entrance pupil radiation brightness to obtain retrieved parameters corresponding to the parameters to be retrieved.

[0072] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0073] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0074] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A collaborative inversion method for atmospheric and oceanic parameters in nearshore waters, characterized by: include: Obtaining parameters to be inverted, the parameters to be inverted including atmospheric parameters and ocean parameters, the ocean parameters including chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter at a wavelength of 440 nanometers, and backscattering coefficient of particulate matter at a wavelength of 660 nanometers; the atmospheric parameters including effective radius, effective variance, particle number concentration, composition information, and complex refractive index of atmospheric aerosols; Calling a multi-parameter physical information neural network model to perform ocean water-leaving radiation calculation on the ocean parameters to obtain ocean water-leaving radiation, and calling an atmospheric aerosol model to perform parameterized calculation on the atmospheric parameters to obtain scattering parameters and absorption coefficients of atmospheric aerosols; Inputting the ocean water-leaving radiation, the scattering parameters and the absorption coefficient of the atmospheric aerosol into a forward radiation transfer model, performing radiation transfer calculation, and obtaining the entrance pupil radiation brightness at the top of the atmosphere; Parameter inversion is performed based on the satellite measurement radiation signal and the entrance pupil radiation brightness to obtain inversion parameters corresponding to the parameters to be inverted.

2. The method for collaborative inversion of atmospheric and oceanic parameters in nearshore waters according to claim 1, characterized in that: The multi-parameter physical information neural network model is obtained by training a training sample set, and the construction process of the training sample set includes: Construct ocean parameter samples; Normalizing the ocean parameter samples to obtain input data of a training sample set; Calling a preset marine bio-optical model to calculate the ocean water-leaving radiation on the ocean parameter sample to obtain a data sample of the ocean water-leaving radiation; At each wavelength in the light wavelength range, performing a Fourier expansion calculation on the data sample at a relative azimuth angle to obtain a Fourier coefficient corresponding to the wavelength, wherein the relative azimuth angle is the difference between the solar azimuth angle and the satellite azimuth angle, and the light wavelength range is 250 nanometers to 800 nanometers, with a wavelength interval of 5 nanometers; The Fourier coefficients are subjected to dimensionality reduction processing by principal component analysis, and the obtained Fourier coefficients corresponding to the principal component components are used as output data of the training sample set.

3. The method for collaborative inversion of atmospheric and oceanic parameters in nearshore waters according to claim 2, characterized in that: The process of Fourier coefficient dimensionality reduction processing includes: Perform principal component analysis on the Fourier coefficients of each wavelength to obtain candidate principal component components; Reconstructing the eigenvectors obtained by calculating the candidate principal component components to obtain reconstructed data; Calculating the relative error percentage at each wavelength based on the Fourier coefficients and the reconstructed data; When the average value of the relative error percentage is less than 0.1%, the candidate principal component is used as the principal component after dimensionality reduction.

4. The method for collaborative inversion of atmospheric and oceanic parameters in nearshore waters according to claim 2, characterized in that: In the ocean parameter sample, the absorption coefficient of particulate matter at a wavelength of 440 nanometers satisfies the following constraints: The backscattering coefficient of particles under light with a wavelength of 660 nanometers satisfies the following constraints: in, It represents the absorption coefficient of particles at a wavelength of 440 nanometers. represents the absorption coefficient of phytoplankton at a wavelength of 440 nm. It represents the backscattering coefficient of particles at a wavelength of 660 nanometers. represents the absorption coefficient of phytoplankton at a wavelength of 660 nanometers. It represents the absorption coefficient of particles at a wavelength of 660 nanometers.

5. The method for collaborative inversion of atmospheric and oceanic parameters in nearshore waters according to claim 1, characterized in that: The training process of the multi-parameter physical information neural network model includes: Constructing a root mean square error loss function based on the input data of the training sample set and the output data of the training sample set; Back propagation is performed in the multi-parameter physical information neural network model according to the root mean square error loss function to update the parameters of the multi-parameter physical information neural network model.

6. The method for collaborative inversion of atmospheric and oceanic parameters in nearshore waters according to claim 1, characterized in that: The performing parameter inversion according to the satellite measured radiation signal and the entrance pupil radiation brightness to obtain the inversion parameter corresponding to the parameter to be inverted includes: Determining an error value between the satellite measured radiation signal and the entrance pupil radiation brightness; Constructing an inversion cost function for the state vector of the parameter to be inverted according to the error value; The minimum value of the inversion cost function is calculated, and the state vector of the parameter to be inverted when the inversion cost function obtains the minimum value is determined as the inversion parameter corresponding to the parameter to be inverted.

7. A collaborative inversion system for atmospheric and oceanic parameters in nearshore waters, characterized by: include: An acquisition module is used to acquire parameters to be inverted, wherein the parameters to be inverted include atmospheric parameters and ocean parameters. The ocean parameters include chlorophyll concentration, solar zenith angle, satellite observation zenith angle, absorption coefficient of particulate matter at a wavelength of 440 nanometers, and backscattering coefficient of particulate matter at a wavelength of 660 nanometers. The atmospheric parameters include effective radius, effective variance, particle number concentration, composition information, and complex refractive index of atmospheric aerosols. a calculation module, configured to call a multi-parameter physical information neural network model to perform ocean water-leaving radiation calculation on the ocean parameters to obtain ocean water-leaving radiation, and call an atmospheric aerosol model to perform parameterized calculation on the atmospheric parameters to obtain scattering parameters and absorption coefficients of atmospheric aerosols; a transmission module, configured to input the ocean water-leaving radiation, the scattering parameters and the absorption coefficient of the atmospheric aerosol into a forward radiation transmission model, perform radiation transmission calculations, and obtain the entrance pupil radiation brightness at the top of the atmosphere; The inversion module is used to perform parameter inversion based on the satellite measurement radiation signal and the entrance pupil radiation brightness to obtain the inversion parameters corresponding to the parameters to be inverted.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method for collaborative inversion of nearshore atmospheric and oceanic parameters as claimed in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for collaborative inversion of atmospheric and oceanic parameters in nearshore waters as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for collaborative inversion of atmospheric and oceanic parameters in nearshore waters as claimed in any one of claims 1 to 6 is implemented.

Images