A method and system for carbon dioxide inversion based on satellite remote sensing data

Abstract

The present invention provides a method and system for carbon dioxide inversion based on satellite remote sensing data. The method includes the following steps: collecting remote sensing data of atmospheric carbon dioxide based on high-resolution satellites; performing radiometric calibration and atmospheric correction on the remote sensing data to obtain first remote sensing data; establishing a spectral analysis model and performing spectral analysis on the first remote sensing data to obtain spectral characteristic information; establishing an interference filtering model and performing interference filtering on the first remote sensing data to obtain second remote sensing data; constructing a carbon dioxide inversion model based on the spectral characteristic information and the second remote sensing data; and converting the spectral characteristic information into carbon dioxide concentration information based on the carbon dioxide inversion model to complete the carbon dioxide inversion based on satellite remote sensing data.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing monitoring, and particularly relates to a carbon dioxide inversion method and system based on satellite remote sensing data. Background Art

[0002] Climate change caused by the greenhouse effect has always been a hot issue of international concern. As the main greenhouse gas in the atmosphere, carbon dioxide is an important factor causing the greenhouse effect. In order to study the distribution of atmospheric carbon dioxide and the changing law of carbon dioxide on a global scale, and to formulate reasonable carbon dioxide emission reduction policies, high-precision carbon dioxide detection data is required. Detection methods based on ground stations and near-earth aircraft have high measurement accuracy, but they all have problems such as narrow detection range and insufficient data volume. Therefore, both ground-based and air-based observation methods cannot meet the measurement requirements of atmospheric carbon dioxide for global climate change research. With the rapid development of remote sensing satellite technology, satellite remote sensing methods based on space-based measurements can achieve high-frequency global-scale detection. At present, many remote sensing satellites for atmospheric greenhouse gases have been launched internationally for greenhouse gas monitoring and climate change research. However, in order to meet the actual application requirements, the accuracy of satellite remote sensing carbon dioxide inversion must be higher than 1%. High-precision carbon dioxide inversion is one of the key factors for realizing the application of remote sensing satellites. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides a carbon dioxide inversion method and system based on satellite remote sensing data, and obtains more accurate carbon dioxide inversion results through spectral processing and analysis.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A carbon dioxide inversion method based on satellite remote sensing data includes the following steps:

[0006] Collect remote sensing data of atmospheric carbon dioxide based on high-resolution satellites;

[0007] Perform radiometric calibration and atmospheric correction on the remote sensing data to obtain first remote sensing data;

[0008] Establish a spectral analysis model, perform spectral analysis on the first remote sensing data, and obtain spectral characteristic information;

[0009] Establish an interference filtering model, perform interference filtering on the first remote sensing data, and obtain second remote sensing data;

[0010] Construct a carbon dioxide inversion model based on the spectral characteristic information and the second remote sensing data;

[0011] Based on the carbon dioxide inversion model, convert the spectral feature information into carbon dioxide concentration information to complete the carbon dioxide inversion based on satellite remote sensing data.

[0012] Preferably, the method for radiometric calibration of the remote sensing data is as follows:

[0013] Convert the radiation information of carbon dioxide collected by the high-resolution satellite into grayscale values to obtain an uncalibrated remote sensing grayscale image;

[0014] Based on the metadata file when the high-resolution satellite collects the radiation information of carbon dioxide, obtain the offset coefficient and gain coefficient;

[0015] Based on the offset coefficient and the gain coefficient, establish a linear equation, and based on the linear equation, convert the uncalibrated remote sensing grayscale image into a radiance value image to complete the radiometric calibration of the remote sensing data.

[0016] Preferably, the method for atmospheric correction of the remote sensing data is as follows:

[0017] Based on the synchronous atmospheric corrector, collect the raw atmospheric data;

[0018] Based on the radiometric calibration, calculate the polarization information of the raw atmospheric data to obtain geometric data;

[0019] Based on the geometric data, identify and remove the cloud pixels in the raw atmospheric data to obtain the initial atmospheric data;

[0020] Based on the geometric data and the polarization contribution of the atmospheric path radiation, calculate the apparent polarization reflectance;

[0021] Based on the apparent polarization reflectance, polarization channels, and polarization radiance, construct a cost function; based on the cost function, obtain the aerosol inversion result;

[0022] Based on the surface reflectance and the total transmittance of the atmosphere in the observation direction, obtain the relationship between the apparent reflectance at the top of the atmosphere and the surface reflectance;

[0023] Based on the preset band atmospheric path radiation and the relationship between the apparent reflectance at the top of the atmosphere and the surface reflectance, obtain the water vapor absorption transmittance;

[0024] Based on the water vapor absorption transmittance and the prior fitting coefficient, obtain the water vapor column content; based on the water vapor column content, obtain the water vapor inversion result;

[0025] Integrate the initial atmospheric data, the aerosol inversion result, and the water vapor inversion result to obtain the synchronous atmospheric parameters;

[0026] Based on the synchronous atmospheric parameters, complete the atmospheric correction of the remote sensing data.

[0027] Preferably, the method for calculating the geometric data by resolving the polarization information of the original atmospheric data is as follows:

[0028] Based on the radiation calibration, calculate the target Stokes parameters of the polarization channel:

[0029]

[0030] where I λ is the total intensity in the λ band, Q λ and U λ are both the linear polarization degrees in the λ band; A λ is the offset coefficient, Z λ is the gain coefficient, is the polarization azimuth angle; DN λ is the gray value, T λ is the time;

[0031] Based on the target Stokes parameters, calculate the polarization degree;

[0032] Based on the polarization degree, obtain the geometric data.

[0033] Preferably, the method for establishing a spectral analysis model to perform spectral analysis on the first remote sensing data and obtain spectral feature information is as follows:

[0034] Reduce the dimension of the first remote sensing data, and obtain the fitness function based on the amplitude gain of the dimension-reduced first remote sensing data and the hidden layer activation function;

[0035] Perform pixel recombination on the dimension-reduced first remote sensing data, and obtain the regional pixel distribution based on the fitness function;

[0036] Based on the regional pixel distribution and the data sparse feature components, obtain the edge pixel level of the spectral feature information;

[0037] Based on the edge pixel level, construct a blur detection model to perform information detection on the first remote sensing data, and obtain the autocorrelation statistical feature quantity of the first remote sensing data;

[0038] Based on the autocorrelation statistical feature quantity, use the fuzzy information mining algorithm to extract features from the first remote sensing data, and obtain the primary spectral feature information;

[0039] Fuse the obtained primary spectral feature information, filter the fused primary spectral feature information, and use the correlation feature detection method to obtain the spectral feature distribution;

[0040] Based on the spectral feature distribution, obtain the final spectral feature information of the first remote sensing data.

[0041] Preferably, the method for establishing the interference filtering model is as follows:

[0042] Add noise to the first remote sensing data, perform empirical mode decomposition on the first remote sensing data after adding noise, and obtain the intrinsic mode function components;

[0043] Based on the correlation coefficient between the intrinsic mode function components and the noise, obtain the noisy intrinsic mode function components;

[0044] Filter the noisy intrinsic mode function components and reconstruct them with the remaining intrinsic mode function components to complete the construction of the interference filtering model.

[0045] The present invention also provides a carbon dioxide inversion system based on satellite remote sensing data for implementing the carbon dioxide inversion method, including: a collection module, a calibration and correction module, a spectral analysis model construction module, an interference filtering model construction module, an inversion model construction module, and an inversion module;

[0046] The collection module is used to collect atmospheric carbon dioxide remote sensing data based on high-resolution satellites;

[0047] The calibration and correction module is used to perform radiometric calibration and atmospheric correction on the remote sensing data to obtain the first remote sensing data;

[0048] The spectral analysis model construction module is used to establish a spectral analysis model, perform spectral analysis on the first remote sensing data, and obtain spectral feature information;

[0049] The interference filtering model construction module is used to establish an interference filtering model, perform interference filtering on the first remote sensing data, and obtain the second remote sensing data;

[0050] The inversion model construction module is used to construct a carbon dioxide inversion model based on the spectral feature information and the second remote sensing data;

[0051] The inversion module is used to convert the spectral feature information into carbon dioxide concentration information based on the carbon dioxide inversion model to complete the carbon dioxide inversion based on satellite remote sensing data.

[0052] Preferably, the calibration and correction module includes a radiometric calibration unit, and the radiometric calibration unit includes:

[0053] A grayscale conversion sub-unit is used to convert the radiation information of carbon dioxide collected by the high-resolution satellite into a grayscale value to obtain an uncalibrated remote sensing grayscale image;

[0054] A coefficient acquisition subunit, configured to obtain an offset coefficient and a gain coefficient based on a metadata file when a high-resolution satellite acquires radiation information of carbon dioxide;

[0055] A radiation calibration subunit, configured to establish a linear equation based on the offset coefficient and the gain coefficient, and convert the uncalibrated remote sensing grayscale image into a radiance value image based on the linear equation, thereby completing the radiation calibration of the remote sensing data.

[0056] Preferably, the calibration and correction module further includes an atmospheric correction unit, and the atmospheric correction unit includes:

[0057] A geometric solution operator unit, configured to acquire raw atmospheric data based on a synchronous atmospheric corrector; and perform solution on polarization information of the raw atmospheric data based on the radiation calibration to obtain geometric data;

[0058] A cloud pixel elimination subunit, configured to identify and eliminate cloud pixels in the raw atmospheric data based on the geometric data to obtain initial atmospheric data;

[0059] An aerosol inversion subunit, configured to calculate an apparent polarization reflectance based on the geometric data and the polarization contribution of atmospheric path radiation; construct a cost function based on the apparent polarization reflectance, a polarization channel, and a polarized radiance; and obtain an aerosol inversion result based on the cost function;

[0060] A water vapor inversion subunit, configured to obtain a relationship between the reflectance at the top of the atmosphere and the surface reflectance based on the surface reflectance and the total transmittance of the atmosphere in the observation direction; obtain a water vapor absorption transmittance based on the preset band atmospheric path radiation and the relationship between the reflectance at the top of the atmosphere and the surface reflectance; obtain a water vapor column content based on the water vapor absorption transmittance and a prior fitting coefficient; and obtain a water vapor inversion result based on the water vapor column content;

[0061] An atmospheric correction subunit, configured to integrate the initial atmospheric data, the aerosol inversion result, and the water vapor inversion result to obtain synchronous atmospheric parameters; and complete the atmospheric correction of the remote sensing data based on the synchronous atmospheric parameters.

[0062] Preferably, the spectral analysis model construction module includes:

[0063] A dimensionality reduction unit, configured to perform dimensionality reduction on the first remote sensing data, and obtain a fitness function based on the amplitude gain of the dimensionally reduced first remote sensing data and a hidden layer activation function;

[0064] A pixel distribution acquisition unit, configured to perform pixel recombination on the dimensionally reduced first remote sensing data, and obtain a regional pixel distribution based on the fitness function;

[0065] An edge pixel-level acquisition unit, configured to obtain the edge pixel-level of spectral feature information based on the regional pixel distribution and the data sparse feature components;

[0066] A primary spectral feature information acquisition unit, configured to build a blurriness detection model based on the edge pixel-level, perform information detection on the first remote sensing data, and obtain the autocorrelation statistical feature quantity of the first remote sensing data; based on the autocorrelation statistical feature quantity, adopt a fuzzy information mining algorithm to extract features from the first remote sensing data to obtain primary spectral feature information;

[0067] An information fusion unit, configured to fuse the obtained primary spectral feature information, filter the fused primary spectral feature information, and obtain the spectral feature distribution by using a correlation feature detection method; based on the spectral feature distribution, obtain the final spectral feature information of the first remote sensing data.

[0068] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention performs radiometric calibration and atmospheric correction on remote sensing data to obtain the first remote sensing data; uses the method of synchronous atmospheric correction to preprocess the remote sensing data and improve the accuracy of the inversion result. A spectral analysis model is established to perform spectral analysis on the first remote sensing data to obtain spectral feature information; an interference filtering model is established to filter interference from the first remote sensing data to obtain the second remote sensing data; the present invention performs denoising and filtering on the remote sensing data through a multi-modal decomposition method, reducing signal interference. Based on the spectral feature information and the second remote sensing data, a carbon dioxide inversion model is constructed; through the inversion of the carbon dioxide concentration, its distribution characteristics are obtained, providing technical support for atmospheric monitoring and environmental governance. The present invention adopts a more comprehensive remote sensing data calibration and correction method to comprehensively improve the inversion accuracy. Description of the Drawings

[0069] In order to more clearly illustrate the technical solutions of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0070] Figure 1 It is a flowchart of a carbon dioxide inversion method based on satellite remote sensing data according to an embodiment of the present invention;

[0071] Figure 2 It is the surface reflectance after synchronous atmospheric correction according to an embodiment of the present invention;

[0072] Figure 3 It is a structural diagram of a carbon dioxide inversion system based on satellite remote sensing data according to an embodiment of the present invention. Detailed Embodiments

[0073] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0074] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0075] Embodiment 1

[0076] As Figure 1 shown, a method for retrieving carbon dioxide based on satellite remote sensing data includes the following steps:

[0077] S1: Based on high-resolution satellites, collect remote sensing data of atmospheric carbon dioxide; to construct a carbon dioxide retrieval model, corresponding temperature data, humidity data, pressure, and atmospheric parameters are also collected in this embodiment.

[0078] S2: Perform radiometric calibration and atmospheric correction on the remote sensing data to obtain the first remote sensing data.

[0079] A further implementation method is that the method for performing radiometric calibration on the remote sensing data is:

[0080] Convert the radiation information of carbon dioxide collected by high-resolution satellites into grayscale values to obtain an uncalibrated remote sensing grayscale image;

[0081] Based on the metadata file when high-resolution satellites collect the radiation information of carbon dioxide, obtain the offset coefficient and gain coefficient.

[0082] Based on the offset coefficient and gain coefficient, establish a linear equation, and based on the linear equation, convert the uncalibrated remote sensing grayscale image into a radiance value image to complete the radiometric calibration of the remote sensing data.

[0083] A further implementation method is that the method for performing atmospheric correction on the remote sensing data is

[0084] Based on a synchronous atmospheric corrector, collect raw atmospheric data.

[0085] Based on radiometric calibration, resolve the polarization information of the raw atmospheric data to obtain geometric data; a further implementation method is that the method for resolving the polarization information of the raw atmospheric data to obtain geometric data is:

[0086] Based on radiometric calibration, calculate the target Stokes parameters of the polarization channel:

[0087]

[0088] Among them, I λ is the total intensity in the λ band, Q λ and U λ are both the linear polarization degrees in the λ band; A λ is the offset coefficient, Z λ is the gain coefficient, is the polarization azimuth angle; DN λ is the gray value, T λ is the time.

[0089] Based on the target Stokes parameters, calculate the polarization degree.

[0090] Based on the polarization degree, obtain geometric data. The geometric data includes satellite sensor observation geometry, longitude and latitude, solar observation zenith angle, and azimuth angle.

[0091] Based on the geometric data, identify and remove cloud pixels in the original atmospheric data to obtain initial atmospheric data.

[0092] Based on the geometric data and the polarization contribution of atmospheric path radiance, calculate the apparent polarized reflectance; the formula is as follows:

[0093]

[0094] In the formula, θ s , θ v and Δφ are the solar observation angle, satellite observation angle, and relative azimuth angle; is the polarization contribution of atmospheric path radiance, which is related to the aerosol type and size; is the surface polarized reflectance; and are the transmittances in the upward and downward directions respectively.

[0095] Based on the apparent polarized reflectance, polarization channels, and polarized radiance, construct a cost function; based on the cost function, obtain the aerosol inversion result; in this embodiment, the constructed cost function is as follows:

[0096]

[0097] In the formula, m is the polarization channel, d is the total number of polarization channels, L simu is the simulated normalized polarized radiance, L meas is the actually observed polarized radiance, C is the parameter in the BRDF model, and FMFV is the ratio of the selected volume column concentration to the fine mode of the volume column concentration. In this embodiment, select a set of (AOD, C, FMFV) that minimizes the cost function value as the optimal parameters, and the corresponding AOD is the aerosol inversion result.

[0098] As Figure 2 shown, FLAASH is an existing traditional atmospheric correction method. Based on the surface reflectance and the total transmittance of the atmosphere in the observation direction, the relationship between the apparent reflectance at the top of the atmosphere and the surface reflectance is obtained.

[0099] Based on the atmospheric path radiance in the preset band and the relationship between the apparent reflectance at the top of the atmosphere and the surface reflectance, the water vapor absorption transmittance is obtained.

[0100] Based on the water vapor absorption transmittance and the prior fitting coefficient, the water vapor column content is obtained; based on the water vapor column content, the water vapor inversion result is obtained.

[0101] Integrate the initial atmospheric data, the aerosol inversion result, and the water vapor inversion result to obtain the synchronous atmospheric parameters.

[0102] Based on the synchronous atmospheric parameters, the atmospheric correction of the remote sensing data is completed.

[0103] S3: Establish a spectral analysis model, perform spectral analysis on the first remote sensing data, and obtain spectral feature information.

[0104] A further implementation method is that the method of establishing a spectral analysis model and performing spectral analysis on the first remote sensing data to obtain spectral feature information is as follows:

[0105] Reduce the dimension of the first remote sensing data, and based on the amplitude gain of the dimension-reduced first remote sensing data and the hidden layer activation function, obtain the fitness function.

[0106] Perform pixel recombination on the dimension-reduced first remote sensing data, and based on the fitness function, obtain the regional pixel distribution.

[0107] Based on the regional pixel distribution and the data sparse feature components, obtain the edge pixel level of the spectral feature information:

[0108]

[0109] In the formula, η represents the spatial distribution pixel gain, φ represents the sparse feature component of the first remote sensing data, R is the template matching coefficient, D is the iteration coefficient, and R = D / 2.

[0110] Based on the edge pixel level, construct a blurriness detection model, perform information detection on the first remote sensing data, and obtain the autocorrelation statistical feature quantity of the first remote sensing data.

[0111] Based on the autocorrelation statistical feature quantity, a fuzzy information mining algorithm is used to extract features from the first remote sensing data to obtain primary spectral feature information; in this embodiment, based on the autocorrelation statistical feature quantity, and by analyzing the state space of the first remote sensing data, the high-resolution features of the first remote sensing data are obtained, based on the pixel feature equalization configuration, wavelet decomposition is performed on the high-resolution features, and based on the wavelet decomposition result, the frame distribution sequence of the first remote sensing data is obtained.

[0112] Based on the frame distribution sequence, a kernel function model is established according to the concentration of the first remote sensing data distribution, with the minimum utility threshold as the weighted vector to obtain the neighborhood of the first remote sensing data distribution; based on the neighborhood, the characteristic scattering distribution is obtained, multiple scans and regional information fusion are performed to obtain the clustering center of the first remote sensing data, and based on the fuzzy degree detection method, the weighted utility value of the first remote sensing data is obtained.

[0113] Based on the weighted utility value and the fuzzy membership function of the first remote sensing data, the maximum cross probability of the first remote sensing data is obtained, and the primary spectral feature information is obtained based on the maximum cross probability.

[0114] The obtained primary spectral feature information is fused, and the fused primary spectral feature information is filtered, and the spectral feature distribution is obtained by using the correlation feature detection method.

[0115] Based on the spectral feature distribution, the final spectral feature information of the first remote sensing data is obtained.

[0116] S4: Establish an interference filtering model to filter the interference of the first remote sensing data to obtain the second remote sensing data;

[0117] A further implementation manner lies in that the method for establishing the interference filtering model is as follows:

[0118] Noise is added to the first remote sensing data, and empirical mode decomposition is performed on the first remote sensing data after adding noise to obtain the intrinsic mode function components.

[0119] Based on the correlation coefficient between the intrinsic mode function component and the noise, the noisy intrinsic mode function component is obtained.

[0120] The noisy intrinsic mode function component is filtered and reconstructed with the remaining intrinsic mode function components to complete the construction of the interference filtering model.

[0121] S5: Based on the spectral feature information and the second remote sensing data, construct a carbon dioxide inversion model. In this embodiment, a feedforward neural network is used to construct the carbon dioxide inversion model. In a feedforward neural network, each neuron is connected to every neuron in the previous layer, and each connection has a weight. The weight value determines the influence degree of each neuron on the signal. Each neuron receives the signal and weight from the previous layer, sums them up after weighting, and is activated by a non-linear function (such as the sigmoid function) to generate an output value, which will be transmitted to all neurons in the next layer as input.

[0122] In this embodiment, the spectral feature information and the second remote sensing data are used as input vectors. After the network processes the data, it outputs a predicted value of the carbon dioxide concentration distribution feature and compares it with the actual value. During the training of the neural network, it is necessary to adjust the connection weights between neurons to improve the performance and accuracy of the network. This process can be regarded as an optimization problem, and it is necessary to find the optimal connection weights so that the output result of the network can be closest to the expected result. Using the quantized conjugate gradient method, the solution of the linear equations is converted into an energy optimization problem, and the parameters with the minimum energy are searched for, so that the output result of the network gradually approaches the true result and obtains the optimal prediction effect of the carbon dioxide concentration distribution feature.

[0123] S6: Based on the carbon dioxide inversion model, convert the spectral feature information into carbon dioxide concentration information (carbon dioxide concentration distribution feature) to complete the carbon dioxide inversion based on satellite remote sensing data.

[0124] Embodiment 2

[0125] As Figure 3 shown, the present invention also provides a carbon dioxide inversion system based on satellite remote sensing data for implementing the carbon dioxide inversion method, including: a collection module, a calibration and correction module, a spectral analysis model construction module, an interference filtering model construction module, an inversion model construction module, and an inversion module;

[0126] The collection module is used to collect atmospheric carbon dioxide remote sensing data based on high-resolution satellites.

[0127] The calibration and correction module is used to perform radiometric calibration and atmospheric correction on the remote sensing data to obtain the first remote sensing data.

[0128] The spectral analysis model construction module is used to establish a spectral analysis model and perform spectral analysis on the first remote sensing data to obtain spectral feature information.

[0129] The interference filtering model construction module is used to establish an interference filtering model and perform interference filtering on the first remote sensing data to obtain the second remote sensing data.

[0130] An inversion model construction module, configured to construct a carbon dioxide inversion model based on spectral feature information and second remote sensing data.

[0131] An inversion module, configured to convert spectral feature information into carbon dioxide concentration information based on the carbon dioxide inversion model, and complete the carbon dioxide inversion based on satellite remote sensing data.

[0132] A further implementation manner lies in that the calibration and correction module includes a radiation calibration unit, and the radiation calibration unit includes:

[0133] A grayscale conversion sub-unit, configured to convert the radiation information of carbon dioxide collected by a high-resolution satellite into grayscale values to obtain an uncalibrated remote sensing grayscale image.

[0134] A coefficient acquisition sub-unit, configured to obtain an offset coefficient and a gain coefficient based on the metadata file when the high-resolution satellite collects the radiation information of carbon dioxide.

[0135] A radiation calibration sub-unit, configured to establish a linear equation based on the offset coefficient and the gain coefficient, and convert the uncalibrated remote sensing grayscale image into a radiance value image based on the linear equation to complete the radiation calibration of the remote sensing data.

[0136] A further implementation manner lies in that the calibration and correction module further includes an atmospheric correction unit, and the atmospheric correction unit includes:

[0137] A geometric solution operator unit, configured to collect raw atmospheric data based on a synchronous atmospheric corrector; and perform solution on the polarization information of the raw atmospheric data based on radiation calibration to obtain geometric data.

[0138] A cloud pixel elimination sub-unit, configured to identify and eliminate cloud pixels in the raw atmospheric data based on the geometric data to obtain initial atmospheric data.

[0139] An aerosol inversion sub-unit, configured to calculate the apparent polarization reflectance based on the geometric data and the polarization contribution of the atmospheric path radiation; construct a cost function based on the apparent polarization reflectance, the polarization channel, and the polarization radiance; and obtain an aerosol inversion result based on the cost function.

[0140] A water vapor inversion sub-unit, configured to obtain the relationship between the apparent reflectance at the top of the atmosphere and the surface reflectance based on the surface reflectance and the total transmittance of the atmosphere in the observation direction; obtain the water vapor absorption transmittance based on the preset band atmospheric path radiation and the relationship between the apparent reflectance at the top of the atmosphere and the surface reflectance; obtain the water vapor column content based on the water vapor absorption transmittance and the prior fitting coefficient; and obtain a water vapor inversion result based on the water vapor column content.

[0141] An atmospheric correction subunit, which is used to integrate initial atmospheric data, aerosol inversion results, and water vapor inversion results to obtain synchronized atmospheric parameters; and based on the synchronized atmospheric parameters, complete the atmospheric correction of remote sensing data.

[0142] A further implementation manner is that the spectral analysis model construction module includes:

[0143] A dimensionality reduction unit, which is used to reduce the dimensionality of the first remote sensing data, and based on the amplitude gain and the hidden layer activation function of the dimension-reduced first remote sensing data, obtain a fitness function.

[0144] A pixel distribution acquisition unit, which is used to perform pixel recombination on the dimension-reduced first remote sensing data, and based on the fitness function, obtain a regional pixel distribution.

[0145] An edge pixel level acquisition unit, which is used to obtain the edge pixel level of spectral feature information based on the regional pixel distribution and the data sparse feature component.

[0146] A primary spectral feature information acquisition unit, which is used to construct a blurriness detection model based on the edge pixel level, perform information detection on the first remote sensing data, and obtain the autocorrelation statistical feature quantity of the first remote sensing data; based on the autocorrelation statistical feature quantity, use a fuzzy information mining algorithm to extract features from the first remote sensing data to obtain primary spectral feature information.

[0147] An information fusion unit, which is used to fuse the obtained primary spectral feature information, filter the fused primary spectral feature information, and use a correlation feature detection method to obtain a spectral feature distribution; based on the spectral feature distribution, obtain the final spectral feature information of the first remote sensing data.

[0148] The above-described embodiments are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for retrieving carbon dioxide based on satellite remote sensing data, characterized in that, It includes the following steps: Collect remote sensing data of atmospheric carbon dioxide based on high-resolution satellites; Perform radiometric calibration and atmospheric correction on the remote sensing data to obtain the first remote sensing data; Establish a spectral analysis model, perform spectral analysis on the first remote sensing data, and obtain spectral feature information; Establish an interference filtering model, perform interference filtering on the first remote sensing data, and obtain the second remote sensing data; Construct a carbon dioxide inversion model based on the spectral feature information and the second remote sensing data; Based on the carbon dioxide inversion model, convert the spectral feature information into carbon dioxide concentration information to complete the carbon dioxide inversion based on satellite remote sensing data; The method for establishing a spectral analysis model and performing spectral analysis on the first remote sensing data to obtain spectral feature information is as follows: Reduce the dimension of the first remote sensing data, and based on the amplitude gain and hidden layer activation function of the dimension-reduced first remote sensing data, obtain a fitness function; where the hidden layer is one or more layers of neurons in a neural network; Perform pixel recombination on the dimension-reduced first remote sensing data, and based on the fitness function, obtain the regional pixel distribution; Based on the regional pixel distribution and data sparse feature components, obtain the edge pixel level of the spectral feature information; Based on the edge pixel level, construct a blurriness detection model, perform information detection on the first remote sensing data, and obtain the autocorrelation statistical feature quantity of the first remote sensing data; Based on the autocorrelation statistical feature quantity, use a fuzzy information mining algorithm to extract features from the first remote sensing data to obtain primary spectral feature information; Fuse the obtained primary spectral feature information, filter the fused primary spectral feature information, and use the correlation feature detection method to obtain the spectral feature distribution; Based on the spectral feature distribution, obtain the final spectral feature information of the first remote sensing data.

2. The carbon dioxide inversion method based on satellite remote sensing data according to claim 1, wherein The method for performing radiometric calibration on the remote sensing data is as follows: Convert the radiation information of carbon dioxide collected by the high-resolution satellite into a grayscale value to obtain an uncalibrated remote sensing grayscale image; Based on the metadata file when the high-resolution satellite collects the radiation information of carbon dioxide, obtain the offset coefficient and gain coefficient; Based on the offset coefficient and gain coefficient, establish a linear equation, and based on the linear equation, convert the uncalibrated remote sensing grayscale image into a radiance value image to complete the radiometric calibration of the remote sensing data.

3. The carbon dioxide inversion method based on satellite remote sensing data according to claim 2, wherein The method for performing atmospheric correction on the remote sensing data is as follows: Collect atmospheric raw data based on a synchronous atmospheric corrector; Based on the radiometric calibration, solve the polarization information of the atmospheric raw data to obtain geometric data; Based on the geometric data, identify and remove cloud pixels in the atmospheric raw data to obtain initial atmospheric data; Based on the geometric data and the polarization contribution of atmospheric path radiance, calculate the apparent polarization reflectance; Based on the apparent polarization reflectance, polarization channels, and polarization radiance, construct a cost function; Based on the cost function, obtain the aerosol inversion result; Based on the surface reflectance and the total transmittance of the atmosphere in the observation direction, obtain the relationship between the apparent reflectance at the top of the atmosphere and the surface reflectance; Based on the relationship between the atmospheric path radiance in a preset band, the apparent reflectance at the top of the atmosphere, and the surface reflectance, the water vapor absorption transmittance is obtained; Based on the water vapor absorption transmittance and the prior fitting coefficients, the water vapor column content is obtained; based on the water vapor column content, the water vapor inversion result is obtained; Integrate the initial atmospheric data, the aerosol inversion result, and the water vapor inversion result to obtain synchronous atmospheric parameters; Based on the synchronous atmospheric parameters, the atmospheric correction of the remote sensing data is completed.

4. The carbon dioxide inversion method based on satellite remote sensing data according to claim 3, wherein The method for calculating the polarization information of the original atmospheric data to obtain geometric data is as follows: Based on the radiation calibration, calculate the target Stokes parameters of the polarization channel: Among them, I λ is the total intensity in the λ band, Q λ and U λ are both the degrees of linear polarization in the λ band; A λ is the offset coefficient, Z λ is the gain coefficient, is the polarization azimuth angle; DN λ is the gray value, T λ is the time; Based on the target Stokes parameters, calculate the degree of polarization; Based on the degree of polarization, obtain the geometric data.

5. The carbon dioxide inversion method based on satellite remote sensing data according to claim 1, wherein The method for establishing the interference filtering model is as follows: Add noise to the first remote sensing data, and perform empirical mode decomposition on the noisy first remote sensing data to obtain intrinsic mode function components; Based on the correlation coefficient between the intrinsic mode function components and the noise, obtain the noisy intrinsic mode function components; Filter the noisy intrinsic mode function components and reconstruct them with the remaining intrinsic mode function components to complete the construction of the interference filtering model.

6. A carbon dioxide inversion system based on satellite remote sensing data, characterized in that, For implementing the carbon dioxide inversion method according to any one of claims 1-5, it includes: an acquisition module, a calibration and correction module, a spectral analysis model construction module, an interference filtering model construction module, an inversion model construction module, and an inversion module; The acquisition module is used to collect atmospheric carbon dioxide remote sensing data based on high-resolution satellites; The calibration and correction module is used to perform radiation calibration and atmospheric correction on the remote sensing data to obtain the first remote sensing data; The spectral analysis model construction module is used to establish a spectral analysis model, perform spectral analysis on the first remote sensing data, and obtain spectral feature information; The interference filtering model construction module is used to establish an interference filtering model, perform interference filtering on the first remote sensing data, and obtain the second remote sensing data; The inversion model construction module is used to construct a carbon dioxide inversion model based on the spectral feature information and the second remote sensing data; The inversion module is used to convert the spectral feature information into carbon dioxide concentration information based on the carbon dioxide inversion model, and complete the carbon dioxide inversion based on satellite remote sensing data.

7. The carbon dioxide inversion system based on satellite remote sensing data according to claim 6, characterized in that, The calibration and correction module includes a radiation calibration unit, and the radiation calibration unit includes: A grayscale conversion sub-unit, which is used to convert the radiation information of carbon dioxide collected by the high-resolution satellite into grayscale values to obtain an uncalibrated remote sensing grayscale image; A coefficient acquisition sub-unit, which is used to obtain the offset coefficient and the gain coefficient based on the metadata file when the high-resolution satellite collects the radiation information of carbon dioxide; A radiation calibration sub-unit, which is used to establish a linear equation based on the offset coefficient and the gain coefficient, and convert the uncalibrated remote sensing grayscale image into a radiation brightness value image based on the linear equation to complete the radiation calibration of the remote sensing data.

8. The carbon dioxide inversion system based on satellite remote sensing data according to claim 6, characterized in that, The calibration and correction module further includes an atmospheric correction unit, and the atmospheric correction unit includes: A geometric solution operator unit for collecting raw atmospheric data based on a synchronous atmospheric corrector; resolving the polarization information of the raw atmospheric data based on the radiation calibration to obtain geometric data; A cloud pixel elimination sub-unit for identifying and eliminating cloud pixels in the raw atmospheric data based on the geometric data to obtain initial atmospheric data; An aerosol inversion sub-unit for calculating the apparent polarization reflectance based on the geometric data and the polarization contribution of the atmospheric path radiation; constructing a cost function based on the apparent polarization reflectance, the polarization channel, and the polarization radiance; obtaining an aerosol inversion result based on the cost function; A water vapor inversion sub-unit for obtaining the relationship between the reflectance at the top of the atmosphere and the surface reflectance based on the surface reflectance and the total transmittance of the atmosphere in the observation direction; obtaining the water vapor absorption transmittance based on the preset band atmospheric path radiation and the relationship between the reflectance at the top of the atmosphere and the surface reflectance; obtaining the water vapor column content based on the water vapor absorption transmittance and the prior fitting coefficient; obtaining a water vapor inversion result based on the water vapor column content; An atmospheric correction sub-unit for integrating the initial atmospheric data, the aerosol inversion result, and the water vapor inversion result to obtain synchronous atmospheric parameters; and completing the atmospheric correction of the remote sensing data based on the synchronous atmospheric parameters.

9. The carbon dioxide inversion system based on satellite remote sensing data according to claim 6, wherein The spectral analysis model construction module includes: A dimensionality reduction unit for reducing the dimensionality of the first remote sensing data, and obtaining a fitness function based on the amplitude gain of the dimensionally reduced first remote sensing data and the hidden layer activation function; A pixel distribution acquisition unit for re-organizing the pixels of the dimensionally reduced first remote sensing data, and obtaining a regional pixel distribution based on the fitness function; An edge pixel level acquisition unit for obtaining the edge pixel level of the spectral feature information based on the regional pixel distribution and the data sparse feature component; A primary spectral feature information acquisition unit for constructing a fuzziness detection model based on the edge pixel level, detecting the information of the first remote sensing data, and obtaining the autocorrelation statistical feature quantity of the first remote sensing data; extracting features from the first remote sensing data using a fuzzy information mining algorithm based on the autocorrelation statistical feature quantity to obtain primary spectral feature information; An information fusion unit for fusing the obtained primary spectral feature information, filtering the fused primary spectral feature information, and obtaining a spectral feature distribution using a correlation feature detection method; obtaining the final spectral feature information of the first remote sensing data based on the spectral feature distribution.

Images